UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Effect of yarding operations on log recovery value Copithorne, Robert Dalgleish 1993

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1994-0076.pdf [ 1.14MB ]
Metadata
JSON: 831-1.0075161.json
JSON-LD: 831-1.0075161-ld.json
RDF/XML (Pretty): 831-1.0075161-rdf.xml
RDF/JSON: 831-1.0075161-rdf.json
Turtle: 831-1.0075161-turtle.txt
N-Triples: 831-1.0075161-rdf-ntriples.txt
Original Record: 831-1.0075161-source.json
Full Text
831-1.0075161-fulltext.txt
Citation
831-1.0075161.ris

Full Text

EFFECT OF YARDING OPERATIONSON LOG RECOVERY VALUEbyROBERT DALGLEISH COPITHORNEBachelor of Science in Forestry,The University of Alberta, 1979A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF FORESTRYinTHE FACULTY OF GRADUATE STUDIESDepartment of ForestryWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIADecember 1993© Robert Dalgleish Copithorne, 1993In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature)Department of S . JThe University of British ColumbiaVancouver, CanadaDate ()&312 2! 93DE-6 (2/88)11ABSTRACTThis thesis deals with the problem of controlling losses in logrecovery value in yarding operations. It is believed that lossesunder current operating methods can be substantial over time.The objective is to find ways of applying statistical qualitycontrol methods to controlling and reducing these losses.A literature review was done to determine the statisticaltechniques available. A pilot study was done concurrently todetermine the current levels of value loss and test methods ofmeasuring value loss. The final step was to recommend aprocedure for monitoring and controlling log value losses basedon the findings of the literature review and the pilot study.Value loss estimates in the pilot study were consistent withlosses discussed in the literature (1-2%). The problemsencountered with the methods used in the pilot study, however,were believed to be severe enough to limit the usefulness of themethods and the results. The methods were found to be cumbersomeand time consuming as well as needing considerable refinement inthe area of scaling and grading for true value recognition.The recommended statistical quality control method can be usedoperationally without the necessity of immediate improvement ofscaling and grading techniques. The method involves observingoccurrences of log damage and recording detailed informationrelated to those occurrences so that the causes of the damage candetermined. Procedures can then be developed to correct thesecauses. The method relies on the use of statistical controliiicharts for attributes for monitoring log damage and highlightingperiods of abnormal results for further investigation.ivTABLE OF CONTENTSABSTRACT iiTABLE OF CONTENTS ivLIST OF TABLES viLIST OF FIGURES viiACKNOWLEDGMENT viiiINTRODUCTION 1CHAPTER ONE - LOG RECOVERY VALUE 2Potential gains: 3Objectives: 5CHAPTER TWO - LITERATURE REVIEW 6Quality control concepts: 7Developing a quality control system: 9CHAPTER THREE- PILOT STUDY 34Methods: 35Results: 42Analysis of Damage: 48CHAPTER FOUR- RECOMMENDED QUALITY CONTROL METHOD: 58Operational quality control procedures: 58Long term application of control charts: 64Analyzing the system: 66CHAPTER FIVE - DISCUSSION: 67VLiterature review: 67Pilot study: 68Recommended quality control system: 69Conclusions: 70BIBLIOGRAPHY 73APPENDIX - MAPS 75viLIST OF TABLESTable 1. Sample data and control limit calculationsfor X-bar chart for variables. 12Table 2. Sample data and control limit calculations forattributes chart with variable sample size. 23Table 3. Site characteristics. 36Table 4. Yarding system characteristics. 37Table 5. Sample statistics. 39Table 6. Loss in recovery value by system. 43Table 7. Yarding results and statistics by period. 45viiLIST OF FIGURESFigure 1. Shewhart X-Bar chart for variables. 11Figure 2. Operating characteristic curves forvarying sample sizes. 19Figure 3. Shewhart chart for attributes. 22Figure 4. Checksheet for damage type (partial). 29Figure 5. Checksheet for damage causes and conditions(partial). 30Figure 6. Control chart for log damage (study data). 46Figure 7. Stabilized control chart for log damage. 47Figure 8. Histogram of damage by type. 49Figure 9. Pareto chart of damage by type. 49Figure 10. Checksheet for yarding damage assessment. 59Figure 11. Control chart for yarding damage. 63viiiACKNOWLEDGMENTThe members of my academic committee who greatly assisted me inpreparation and development of the overall project, the fieldstudy, and the completion of this thesis were: Glen Young, JoeMcNeel and Tom Maness.The funding for this research was provided by the Science Councilof British Columbia.1INTRODUCTIONLosses in log value due to yarding damage can have a significanteffect on the profitability of a harvesting operation. Estimatesof the amount of value lost in the literature are in the order of2 percent of the total value of the logs produced. This canamount to millions of dollars for some larger harvestingoperations. The first step in reducing this type of loss is todetermine the causes of the damage. Measures to reduce damagecan then be developed based on widely used statistical qualitycontrol techniques.This thesis is concerned with finding ways to reduce log damageassociated with the yarding methods commonly used on the coast ofBritish Columbia. To begin the process of gathering datanecessary for development of improved methods, a pilot study wasdone. This is discussed in detail in the methods section. Aliterature review was carried out to determine the state ofcurrent technology and find out what methods are being used inother locations to control harvesting operations. The knowledgegained in the pilot study and the literature review was appliedin the development of a comprehensive method of controlling valueloss in yarding operations. The details of this method areprovided in Chapter 4.2CHAPTER ONE - LOG RECOVERY VALUELog recovery value is the value of the standing tree which isrecovered through logs delivered to the end user, either aconverting mill or the open market. It reflects the value of theproducts which can be produced from the logs and excludes thewaste or unusable portion of the tree. In a sawmill, one of theprime objectives is to recover the highest total value from theraw material, and similarly, all phases of harvesting operationsshould be concerned with recovering the highest value from thestanding tree.Little attention has been paid in the past to evaluating theeffects of yarding operations on the value of the logs recovered.Cost (per unit volume) has been the main criteria for measuringperformance and planning yarding operations.This emphasis on costs ignores the fact that profitability is afunction of both cost and value, as shown by the profit equation(Murphy and Twaddle, 1985):Profit = Volume x (Unit Value - Unit Cost)Losses can occur when inappropriate yarding methods are used, orare used incorrectly, causing log damage and reducing log value.Damage to logs means breakage at any point and or slabbing orsplitting of the butt or top. It also includes other types ofdamage such as bark removal, crushing of the wood fibres bychokers and penetration or ripping of the wood fibres by a3mechanical grapple. Gabrielsson et al (1989) mention the abovetypes of damage, among others, as causing a reduction in thevalue of a sawlog. By reducing any of these types of log damageor breakage, log recovery value can be increased.Potential gains:The potential gains to be achieved through controlling losses aredependent on the amount of the loss in recovery value presentlyoccurring, and how much the losses can be reduced. Losses inrecovery value during the yarding process have been estimated byseveral researchers. These estimates can provide a guide to howmuch value is being lost in a particular system, but the amountof loss reduction that can be achieved can not realistically beestimated without in-depth knowledge of a particular system:• Williston (1979), in estimating the total dollaropportunity that is available through improved practicesin the woods in the Pacific Northwest, stated that 3.5%of the merchantable stem of small trees was lost duringprimary transport and at the landing.• McIntosh (1968) studied losses at various stages of theharvesting process in the interior of British Columbiaand found that breakage during skidding averaged 1.7% ofthe net log scale. McIntosh believes that optimalperformance would result from adequate supervision andnatural pride of the woods crew in “good” operations.• Murphy and Twaddle (1985) summarized studies done in NewZealand in which value losses from extraction and butt4damage were found to be 1 to 2 percent of potentialvalue. It was suggested that statistical quality controltechniques be applied to reducing these losses.No studies were found which applied to the type of site and standconditions existing in the remaining old growth stands on thecoast of British Columbia. Harvesting of old growth stands inthis area is done under some of the most difficult operatingconditions anywhere in the world. Steep slopes, rocky terrain,high rainfall, and extremely varied tree size and age combine toincrease the chance of log damage during yarding. Therefore, itis reasonable to expect that value losses under these conditionswould be in the order of 2% or even more.Extending the above estimate, it is possible to project thepotential dollar loss for a setting, or an entire harvestoperation. Based on a setting size of 15 hectares, a volume perhectare of 1000 m3 and an average value of $60 per m3, a valueloss of 2.0 percent would result in a total loss of $18,000 for asingle setting, or $1.20 per m3. For a typical woods operationproducing 500,000 m3 per year, this would be equivalent to a lossof $600,000 in one year. A small reduction in the percentage ofthis value loss could have a considerable effect on theprofitability of an operation.5Objectives:The objective of this thesis is to develop a quality controlmethod using control charts to reduce damage to logs in yardingoperations under coastal British Columbia operating conditions.The first part of this thesis is a review of standard methods ofquality control as discussed in the literature. This included areview of concepts in the field of statistical process controlrelevant to the control of yarding operations. The reviewprovided a description of practical methods for controllingquality in yarding operations from which a detailed procedure canbe developed.The second part was a pilot study to develop and test theprocedures available for gathering various types of data onyarding operations. The experience gained in the pilot study wasapplied to the refinement of a practical method of qualitycontrol which can be tested in further field studies. Theresults of the pilot study included a review of the problemsencountered and how they affected the choice of the proposedquality control methods. The results also included estimates ofaverage value loss developed to meet the specific objectives ofthe study.6CHAPTER TWO - LITERATURE REVIEWThe first step in developing a quality control system for yardingoperations is a review of the current technology and selection ofthe appropriate statistical control methods. This includes areview of the types of control charts which could be used andplanning the steps necessary for their application.Many books have been written on quality control techniques usingstatistical methods, including texts by Juran and Gryna (1993),Grant and Leavenworth (1988) and Duncan (1965). These texts werereviewed in part, as well as other publications includingresearch by Gitlow (1987) and Scherkenbach (1986), which outlinedthe application of current developments in the field to specificquality control problems. Other sources of information onquality control were lectures and course notes by Maness (1993),the Quality Control Handbook edited by Juran (1962 edition), thejoint LIRO/FIEA Seminar: Quality issues in harvesting andprocessing; and the IUFRO P.3.07-01 Harvesting and productquality inaugural meeting; both held in New Zealand in 1993.Murphy and Twaddle (1985), discuss the use of quality controlcharts to monitor a log making operation at a landing. Stuart,Grace, and LeBel (1993), describe several examples ofapplications in forest operations where various types of chartsincluding histograms, Pareto charts and Shewhart charts have beenused to monitor production processes and assist in controlling7variability. The relevant concepts described in these sourcesare discussed in the following sections.Quality control concepts:Quality control is often thought to mean the detection ofdefective products of a manufacturing process. This is commonlyknown as Acceptance Sampling, where the objective is to determinewhether to accept or reject a particular lot or product based oninspection of the parts making up that lot. In fact, qualitycontrol is much more than acceptance sampling. Statisticalmethods initially developed by Shewhart in 1924 have made itpossible to change the emphasis to managing and controllingproduction processes with the objective of preventing defectivegoods from being produced (Duncan, 1965). The current approachto quality control has two basic aspects:• Ensuring that all goods produced meet the requirements ofthe appropriate specification. This is done bymonitoring each stage of the production process so thatdeviations from normal operating behaviour are detectedand corrected as they occur.• Continuous improvement of the methods of production sothat the quality of goods produced is constantlyincreased. This is achieved by analysis of theproduction system and continued strengthening of areaswhere quality levels are lowest.The basis for effective quality control methods is theappropriate use of statistics. An understanding of basic8statistical concepts, including the concept of variability, isneeded in order to effectively implement quality control methods.In any manufacturing process, there will be some differences inquality between individual units produced, no matter how accurateor well controlled the process. Variation is the amount by whichthe quality of an individual unit will differ from other units asindicated by the measurement of a specific variable or attribute.Variability is the range of values over which the individualmeasurements occur. Variability is usually defined statisticallyby the standard deviation (a). In common manufacturingsituations, when the system is operating in a normal, consistentmanner, 99.73%, or nearly all of the individual measurements,should be within three standard deviations (±3a) of the mean ofthe individual measurements for a total variation of six timesstandard deviation, assuming that the measurements are normallydistributed. This is commonly known as the “six-sigma (6o) rule”(Maness, 1993). Variability is dependent on the physical factorswhich affect a process. Duncan (1965) discusses two types ofvariation: chance variation and special variation.Chance or common variation is the sum of the effects of the wholecomplex of chance causes. These variations occur at random andfollow the laws of statistics. Little can be done about chancevariation except to revise the process. On the other hand, thereis variation caused by special or assignable causes. Assignablecauses are non random events resulting from changes in the9factors which affect the operating characteristics of theprocess.These types of factors include (Duncan, 1965):• Differences among machines.• Differences among workers.• Differences among materials.• Differences in each of the above over time.• Differences in their relationship to one another.Variation in quality between individual units can be reduced andcontrolled, but it cannot be eliminated. It is a natural aspectof any dynamic system. Management actions to control variationshould be based on the cost/benefit relationship of thoseefforts. It may be relatively inexpensive to reduce variationproduced by assignable causes, but as variation becomes more andmore due to chance causes, the cost of reducing variation willincrease until it eventually exceeds the increase in valueachieved. At this point, reappraisal and application ofdifferent techniques may be required to further improve quality.Developing a quality control system:The successful development of a quality control system depends onhow well the causes of variation in a system or process areunderstood. A system or process is a combination of activitiesoperating in a coordinated manner to accomplish a defined goal orobjective. In the process of yarding logs, this means thecombination of people, materials, equipment, methods, and10environment which produces the desired result: logs delivered tothe landing undamaged. Detailed study of the system is requiredto determine its specific operating characteristics.The implementation of quality control methods in a yardingoperation has two distinct phases; eliminating special causes ofvariation and continuous improvement:Phase I: Eliminating special causes of variation:Eliminating special causes of variation begins with monitoringthe process which is to be controlled using statistical chartsover an extended period of time. The aspect of time is importantas it allows short term fluctuations to be compared to long termtrends. This comparison will disclose whether short termfluctuations represent a change in the operating characteristicsof the system which should be investigated or only naturalvariation in the system which should be allowed to continue aslong as the specifications are being met. It is important tonote that in this phase the objective is to reduce thevariability of the system, that is to reduce the variance ofindividual measurements from the mean. Reductions in the meanlevel of breakage are the subject of Phase II discussed below.System monitoring:The most common method of monitoring a system is using controlcharts, of the type known as Shewhart charts. There are two maintypes of Shewhart charts; charts for monitoring variables and11charts for monitoring attributes. Both types are similar inappearance, differing only in what type of data is plotted.Charts for variables:The main types of charts for variables are X-Bar and R charts.They should be used when some measurable quantity is beingmonitored such as the amount of value lost in yarding. X-Barcharts are discussed below to demonstrate how control charts forvariables are prepared.Figure 1 is a typical X-Bar chart prepared using hypotheticaldata from a series of samples of logs where the percent valuelost for each log has been determined by estimating the value ofeach log before and after yarding.Figure 1. Shewhart X-Bar chart for variables.VALUE - - - MEAN — —h- — UPPER —0——— LOWERLOST (%) CONTROL CONTROLLIMIT LIMITSAMPLE NUMBERAfter Murphy and Twaddle (1985).12In Figure 1, the Y-axis shows the range of the variable beingmeasured (value loss percent). The X—axis shows the samplenumber and follows the chronological order in which the sampleswere taken. The calculations of the control limits are discussedbelow.Table 1. Sample data and control limit calculations for X-barchart for variables.Sample No. 1 2 3 4 5 6 7 8 9 10 11 12Value loss 2 8 3 12 18 9 11 24 9 14 2 1(%)Mean value 9.4loss(%)Centre line 9.4 9.4 9.4 9.4 9.4 9.4 9.4 9.4 9.4 9.4 9.4 9.4Upper 20.2 20.2 20.2 20.2 20.2 20.2 20.2 20.2 20.2 20.2 20.2 20.2controllimitLower 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0controllimitStandard Based on long term 3.6Deviation sample averages:3 * Std. 10.8 10.8 10.8 10.8 10.8 10.8 10.8 10.8 10.8 10.8 10.8 10.8DeviationIn the X-Bar chart, the mean of each sample is plotted andcompared to the grand sample mean, which is plotted as the centreline. The grand sample mean is derived from previousobservations, if these are available. The upper and lowercontrol limits are set so that if the system is under control,nearly all of the sample means will fall between the limits.Initially, this is the mean or centre line plus or minus three13times the standard deviation of the mean, based on previousobservations of the population. The total range between upperand lower control limits is therefore six times the standarddeviation based on measurements of individual items in thepopulation. If the calculated value of the lower control limitis less than zero, the limit is set at zero. The data should betested to ensure that it is normally distributed before usingthis method.The standard deviation used in the calculation of control limitscan be derived in several ways; either by analyzing the data fromthe current sample, from previous data, or set by management tomeet certain objectives (Duncan, 1965). If sufficient historicaldata is available, the standard deviation of individual itemssampled can be calculated as shown in Equation 1 (Duncan, 1965).— 2 Equation 1=________Where:s = standard deviation of the sample,xj = the th value of x= the mean value of the sample, andn = the number of items in the sample.If past data is not available from which to determine thestandard deviation, current sample data can be used. In thiscase the range, the difference between the highest and the lowestvalue in each subgroup, is calculated, and the average range,(R), is calculated for all subgroups in the current data. It is14not appropriate to calculate the standard deviation with Equation1, above, using the current data, because the estimate will notbe an independent measure for detecting extreme variations in themeans (Duncan, 1965). Conversion factors are taken from standardstatistical tables. The calculation of the standard deviationwhen ranges are used is shown by Equation 2 (Duncan, 1965).Equation 2s=—d2Where:d2 = tabulated value for n sample size, and= mean of sample ranges(Ref.: Duncan, 1965; Table Dl, p 908).When samples are taken in groups such as 4 or 5 at a time, as isusually done with variables charts, and the averages of thegroups (X) plotted on the control chart, the standard deviationobtained above must be corrected for the effect of using sampleaverages instead of individual measurements. The effect, asdescribed by the Central Limit Theorem, is that the distributionof the means (X) tends to be normal even though the individualmeasurements are not and the variability of the means isdecreased as sample size increases (Grant and Leavenworth, 1988).Therefore the standard deviation of the population must beadjusted by dividing by the square root of the sample size, to beusable for control charts in which sample averages are plottedinstead of individual measurements. This adjustment is shown byEquation 3 (Oakland and Followell, 1990).15s Equation 3s— =%dIWhere:S= Standard deviation of the sample means.Guidelines exist for choosing the number of samples to be taken.Oakland and Followell (1990), suggest that as a rule of thumb forcontrol charts for variables, at least 100 individualmeasurements should be taken over a period of time. Duncan(1965) suggests that sample sizes of 4 or 5 are close to optimumfor detecting large shifts in the process average, such as ashift of 2 times the standard deviation or more. If shifts assmall as one standard deviation are to be detected, sample sizesof 15 to 20 are more effective. The individual measurements cantherefore be taken in 20 groups of 5 and the results averaged foreach group. If the sample size is revised, the control limitsshould be recalculated as well. Samples should be taken fromhomogeneous populations whenever possible; smaller sample sizeswill make it easier to achieve this (Duncan, 1965).Whether the individual measurements or sample means are used toestimate the population standard deviation, the estimated valuesshould be considered preliminary only and used with caution,until enough time has passed to develop long term averages.To use charts for variables, the results of successivemeasurements are plotted and the resulting patterns analyzed. Apoint outside the control limits indicates that a change hasoccurred in the average value of the output from the system, but16nonrandom behaviour of points within the control limits can alsoindicate that a special cause of variation has occurred. Thevarious patterns and their interpretation are discussed in moredetail under attribute charts, below.Charts for attributes:When it is not possible or practical to measure a variabledirectly, control charts for attributes can be used (Maness,1993). These charts can be used when a yes or no question can beanswered concerning a specific attribute, such as whether a logis damaged or not. The attribute to be controlled should beeasily recognizable. The same analysis of trends can be appliedto this type of chart as used for variables charts.Sample sizes for attribute charts are usually larger than forvariables charts because the proportion of the population havingthe attribute in question is usually very low, often as low as 1percent.There are a number of ways of choosing the sample size and thefrequency of sampling. According to Duncan (1965), the methodssuggested in the literature are guidelines only. A completesolution to the problem of determining sample sizes and frequencyof sampling requires knowledge of the costs involved in notcatching shifts in performance when they occur, costs connectedwith amount and frequency of sampling and the probabilities ofshifts occurring.17The most important consideration in choosing the sample size,according to Duncan (1965), is determining the rational subgroup.It is important that the samples represent homogeneous units suchas the output from a particular shift or machine. Mixing samplesfrom several shifts or machines can make it very difficult tofind reasons for variation. Therefore, although a fixed samplesize can be used, it is more effective to do a completeinspection of the output from a specific shift or machine on aregular basis.Calculating sample size requires a knowledge of the amount ofvariation which it is desired to detect. This is because samplesize and sensitivity of the chart are directly related to eachother. Increasing sample size increases sensitivity and viceversa. A preliminary estimate of the expected variation can beobtained by taking samples of the process output until arelatively stable average is obtained. Juran (1962) suggeststhat the minimum sample size can be calculated based on thepreliminary process average using Equation 4:9-Equation 4n= —pWhere:p = average fraction defective, andn = the minimum sample size.No estimate of sensitivity is required for Juran’s equation,because it is based on a implied sensitivity of 3 times thestandard deviation. As this amount of variation is commonly used18to set control limits in control charts, Juran’s method can beused in many circumstances. For example, if the process averageis 0.02, (2%), the calculated sample size using this equation is441 items.Another method suggested by Duncan (1965) is dependent on theshift desired to be detected and the proportion defective in thepopulation as shown by Equation 5. This method determines thesample size required to have a 50% chance of catching an increasein the average, with a single sample (Duncan, 1965).(i- ) Equation 5d=3j pV nWhere:d = shift in average to be detected, andp = fraction defective in population.For example, if the shift to be detected (d), was set at threetimes the standard deviation, the estimated sample size for apopulation with an estimated process average of 2%, would be 445items, very similar to the estimate provided by Juran’s equation.Closer examination will reveal that Duncan’s equation can beresolved into Juran’s equation when the desired sensitivity isequal to p.The relationship between sample size and sensitivity of thesample is best shown with what is known as an OperatingCharacteristic (OC) curve. OC curves are used in many areaswhere statistics are applied, for example in designing samples19for acceptance sampling. An OC curve for an attribute chartshows the probability of a single sample falling within thecontrol limits when the process fraction defective is currentlyabove or below the mean or centre line on the chart. The curveshows the risk of saying the process is in control at the meanlevel if a sample point falls within the limits when the processis actually operating at a different level. That is, the curveshows the chance of not catching a shift in the process averagein the first sample taken after the shift occurs (Duncan, 1965).Figure 2 shows a series of CC curves for a control chart with anupper limit only. Charts for attributes will frequently have nolower control limit, usually when the process fraction defectiveis very small or when it is not considered important for controlpurposes.Figure 2. operating characteristic curves for varying samplesizes.-a‘‘‘N\\\.--‘s\\\c‘.\cs..s\\\\.--\.S.‘..‘-.I I I I I I1A n100; TJCLO.060—D—n=200; tJCL=0.048• n400; tJCL=0.039-—n=800; IJCL=0.033I I I I I I ITh4-I I I I I I Ioperating Characteristic Curve1.00’ ————__________C____ ____ ____ ____ ____H 0.9—-DI05so4)cO.4 ———————Dci —————— —— ——0.3____—— —————a0.10.0 . . - — — —0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10 0.15 0.20 0.25 0.30Process Fraction Defective (p)I I I I I I.—.-- -20In this example, the long term average process fraction defectiveis estimated to be 0.019 (1.9%), a value corresponding to theaverage log breakage experienced during the pilot study. Thesample sizes range from 100 to 800 items and the upper controllimits are 0.060, 0.048, 0.039, and 0.033 respectively,reflecting the increased precision of the sample with increasingsample size. Items in a yarding operation would be individuallogs yarded. Each curve in Figure 2 is based on the same averageprocess fraction defective, and shows the effect of changingsample size on the probability of not catching a shift at variouslevels of current process fraction defective. For example, ifthe current process fraction defective increased to 0.06, therewould be a 0% chance of failing to detect the shift with a samplesize of 800 items. The probability of failing to detect theshift increases to 45% as sample size decreases to 100 items. If95% probability of detecting a shift to this level, (0.06), wasdesired, the sample size would be approximately 400 items, using1Pa to estimate the probability of catching a shift. 1Pa isknown as the “Power of the Test” (Maness, 1993).The above example shows the relationship between sample size andthe ability to detect changes in the process average. Because ofthe difficulty of obtaining consistent sample sizes in loggingwhere conditions are extremely variable and production can changerapidly from one time period to the next, the best approach is tobase sample size on the optimum subgroup for maximum samplingefficiency and use CC curves to determine the reliability of theactual samples that result.21Sample size affects the spread of the control limits, because thestandard deviation which is used to calculate the control limitsis dependent on sample size. Therefore, if sample sizes varysignificantly, i.e. by more than 25%, separate upper and lowercontrol limits should be calculated for each sample point usingthe individual sample sizes (Maness, 1993). The mean probabilityas calculated from all of the samples should still be used foreach calculation. The calculation of control limits forattribute charts is based on the assumption that when thefraction defective in the population is low; 10% or less, thepopulation being sampled follows the binomial distribution(Maness, 1993). When the fraction defective is higher, thenormal distribution can be assumed (Duncan, 1965). For mostapplications to yarding operations, the binomial distributionwould be used, because the number of logs damaged should notexceed 10% of the total logs yarded. Equation 6 can be used forcalculating the standard deviation of samples taken from apopulation with a binomial distribution. The value of p to beused in this formula is the process fraction defective (Duncan,1965).Ip(1- ) Equation 6s=JV nWhere:s = standard deviation of the sample means.When setting up a chart initially, the assumption is made thatthe process is in control. The average fraction defective for22the first group of samples is then used as an estimate of theprocess fraction defective (Duncan, 1965).Figure 3 gives an example of a Shewhart chart for attributes witha variable sample size, using hypothetical data such as might beobtained from observations of logs damaged during yarding. Table2 shows the sample data used for Figure 3 and the method ofcalculation of the control limits.Figure 3. Shewhart chart for attributes.I LOGS - -0- - MEAN — —k- — UPPER —-—— LOWERDAMAGED CONTROL CONTROLLIMIT LIMIT0.06SAMPLE NUMBERIn Figure 3, upper and lower control limits move closer togetheras sample size increases, reflecting the increased confidence inthe estimate of the mean, and farther apart as sample sizedecreases.23Table 2. Sample data and control limit calculationsattributes chart with variable sample size.Initially the control limits are placed at the center line ormean plus or minus three times the standard deviation which givesa total range of six standard deviations. After some experienceis gained, the control limits may be adjusted to suit theobjectives of management. Warning limits can be usedconcurrently with control limits to indicate a need to collectmore data to check on the possibility of the process being out ofcontrol. Warning limits may be set at one or two times thestandard deviation (lo or 2o limits). Ideally the limits shouldbe set at a level such that exceeding that level indicates that apredetermined course of action should be taken, and thereforeforSample No. 1 2 3 4 5 6 7 8 9 10 11 12Sample size 550 249 301 400 275 255 600 175 244 255 350 277Damaged 8 3 8 17 5 2 0 3 6 13 6 5Proportion 0.014 0.012 0.027 0.042 0.018 0.008 0 0.017 0.025 0.051 0.017 0.018Mean 0.019proportioncentre line 0.019 0.019 0.019 0.019 0.019 0.019 0.019 0.019 0.019 0.019 0.019 0.019(Mean)Upper 0.037 0.046 0.043 0.040 0.044 0.045 0.036 0.051 0.046 0.045 0.041 0.044controllimit.Lower 0.002 0 0 0 0 0 0.003 0 0 0 0 0controllimitconstant 0.413 0.413 0.413 0.413 0.413 0.413 0.413 0.413 0.413 0.413 0.413 0.413root of n 23.45 15.78 17.35 20.00 16.58 15.97 24.49 13.23 15.62 15.97 18.71 16.643*Std. 0.018 0.026 0.024 0.021 0.025 0.026 0.017 0.031 0.026 0.026 0.022 0.025Deviation24should be set at a critical value for that action. Setting thelevel too low will result in unnecessary problem solvingactivity, whereas, if the level is too high, actual problems maynot be detected. As discussed above, control limits are loweredas sample size increases. Holding the control limits at the samelevel while increasing the sample size, has the effect of makingthe OC curve steeper, thereby increasing the sensitivity orability of the chart to detect shifts.As for setting sample size and frequency, setting the level ofcontrol limits requires balancing the cost of investigatingproblem situations with the cost of ignoring problems which goundetected. A detailed analysis of costs should be made, whichis beyond the scope of this thesis.Criteria for judging out of control situations are given byseveral authors. These criteria can be used for any type ofchart. According to Duncan (1965) they are:• One or more points outside the control limits.• One or more points in the vicinity of a warning limit.• A run of 7 or more points up or down or above or belowthe control limit. This will indicate a problem emergingor a sustained shift in the process average.• Cycles or other nonrandom patterns in the data.Experienced operators may be able to relate patterns tochanges in operating conditions.• A run of 2 or 3 points outside of 2o limits.• A run of 4 or 5 points outside of 10 limits.25The above criteria are based on the premise that if no pointsfall outside the control limits and if there is no evidence ofnonrandom variation in the points within the limits such as arun, there is not enough evidence to assume that special causesof variation are acting on the process (Duncan, 1965). Thecriteria for the number of points in a run is based on theprobability of getting a run of that number in a sample takenfrom a population of a given size. For example there is a 5percent chance of getting a run of 7 or more in a population of20 (Hansen, 1963). Duncan (1965) cautions that using multiplecriteria increases the risk of looking for trouble when noneexists.Target for reduction of log damage:There is no accepted standard or target level in the industry fordamage during yarding operations. When beginning a controlprogram the target should be zero breakage. Once a monitoringprocess has been set up, long term averages for particular setsof site, stand and equipment parameters will be developed.Review of these averages will give some indication as to thelevel of breakage to be expected under current conditions.However, for maximum effectiveness of the control chart method,the target should remain at zero, particularly during the secondphase, continuous improvement.26Choosing the type of chart:Experience in previous studies has shown that the original volumeof broken logs or the grade of the broken section is difficult todetermine in yarding operations unless detailed measurements oflogs have been made prior to yarding, because of difficulties infinding the broken piece or pieces (McNeel, 1993). The resultantvalue loss cannot be accurately determined without thisinformation. Making detailed measurements prior to yarding maynot be practical under operating conditions because of the largenumber of logs which would have to be measured in order to ensurea representative number of damaged logs is included in theresults. Control, however, can still be achieved by observingand recording qualitative attributes such as details of thefrequency and nature of log damage. Types of log damage havebeen discussed in Chapter 1. For further details, refer topapers by Gabrielsson et al (1989), McIntosh (1968) and Williston(1979). Because it is relatively easy to tell if a log is brokenor damaged, control charts for attributes can used to control thefrequency of log damage. Log damage can be directly controlled,and value loss indirectly controlled. Reducing log damage willresult in the recovery of volume which would not otherwise berecovered and an improvement in the overall value of logsrecovered, both of which will increase the total log recoveryvalue from the setting.27Data collection:Data should be collected by an observer close enough to theyarding operation to observe when damage occurs and to alsoobserve the conditions which contributed to the damage. It maybe practical after the system has been implemented and acceptancehas been gained of the usefulness of the system to have a memberof the yarding crew record the observations (Maness, 1993).Prior to that time, however, a designated quality control officershould carry out the data collection. With a small amount oftraining, the crew members can become at least partly involved inthe control process through participation in the analyticalaspect of the system. This will result in greatly increasedbenefits from the control program. The key to getting theemployees to willingly participate in the program is to make itclear that the results will not be used in any way to evaluatetheir individual performance. It should be stressed theobjective is improve recovery, not to apportion blame orresponsibility for errors (Scherkenbach, 1986).The type of data to be collected would not be fixed untilconsiderable experience has been gained in using the controlmethods. In the beginning, data collection would include as muchinformation as it is feasible to collect about the type of damageand the conditions or circumstances surrounding the damage. Itis best at this point to avoid attributing the damage to any onefactor, as any attempt to limit the information collected will28create the possibility that some vital piece of information isoverlooked.Analyzing a system:Using control charts to monitor a system is only part of theprocess of controlling special causes of variation and achievingpredictability in a system. Completion of the process involvesan in-depth analysis of the system. The objective is todetermine where the most serious causes of variation occur andfind out how to reduce them. Whereas control charts areprimarily numeric, the tools used for this purpose areanalytical, mainly observation and documentation of criticalareas of a process. Particular emphasis is placed on periodswhen special causes of variation occur as indicated by thecontrol charts. Specific tools used for this method includecheck sheets, histograms and Pareto charts.Check Sheets:Dey (1993) states that check sheets provide a disciplinedapproach to gathering factual information. Check sheets can bedeveloped for any process. The key is to identify the criticalfactors. The check sheet is not necessarily limited bystatistical considerations. It is only necessary to observe aprocess and record what is observed, with particular emphasis oncauses of errors and what action was, or could be, taken tocorrect them.29Figure 4 is an example of a check sheet designed to gatherinformation on the amount and type of log damage which occurs.The information from this check sheet can be used as input to acontrol chart for attributes. A column of the checksheet wouldbe used for each log. Each time a log is observed as damaged, amark would be placed opposite the type or types of damage. Thetotal number of logs in the sample would be tallied using someform of mechanical tally device.Figure 4. Checksheet for damaqe type (partial).SAZ4PLENO. 1 2 3 4 5 6 7 8 9 10 11 12TOTALDAMPGE TYPE:BROKEN TOPBROKEN ENDSHATTERED ENDSLABBEDTOP SLABBEDBROKEN BUTTBUTT SPLITOTHER:TOTAL LOGS YARDEDTOTAL LOGS DAMAGEDFigure 5 is a check sheet focused on the causes of log damage andthe conditions existing at the time of the damage. Since therecan be more than one contributing factor to a damage occurrence,data from this type of checksheet is not usually used forstatistics on numbers of logs damaged, but can be used forfurther analysis of causes of damage.30Figure 5. Checksheet for damage causes and conditions (partial).LOG NO. 1 2 ._i9 ._iYARDING CONDITIONS:BREAKOUTHANGUP ON STUMPGULLEYLANDINGYARDING MACHINEYARDING SYSTEMNG INTERWEATHERSLOPETERRAINSTAND AGECABLE RIGGINGHO OF CHOKERSGRAPPLEDEFLECTIONYARDING DISTANCEOTHER:The checksheet for damage types can be combined with thechecksheet for damage causes and conditions to make one datacollection form which can be used in the field. For an exampleof this combined form, see Chapter 4, recommended quality controlsystem. Once a number of samples has been taken, the data in thedamage type section on the number of logs yarded and damaged canbe transferred to a process control chart for further analysis.The damage causes section of the check sheet will provide detailsnecessary for further investigation of special variation when itoccurs in a sample.Histograms and Pareto charts:Histograms are an often overlooked means of showing graphicallywhich types of problems occur most frequently. They can be usedto determine where problem solving efforts should beconcentrated. In Figure 8 (Chapter 3), a histogram is used toanalyse the frequency of various types of log damage thatoccurred during the pilot study.31Pareto charts are an easily used method of analyzing data. Asmentioned by Juran and Gryna (1993), they are based on the ParetoPrinciple which states that a small number of the causes of lossis often responsible for a large amount of the loss. Figure 9(Chapter 3), is an example of a Pareto chart.With a Pareto chart, causes of variation are ranked according tofrequency of occurrence, effect on cost, or some other relevantbasis. The Pareto chart in Figure 9, shows these causes rankedin order of descending value loss based on the actual datacollected during the pilot study. The chart also shows thecumulative value loss as a result of the causes. Figure 9 showswhich causes have the greatest effect and where efforts can beconcentrated to achieve the greatest improvement. In Figure 9,the cumulative loss line shows that nearly 80% of the value lossis of two types, broken ends and split butts. As experience isgained, the information collected will be refined to suit thecircumstances. The Pareto chart categories can be modified asappropriate.Once a system has been analyzed and special causes of variationinvestigated and removed (where cost effective), the first phaseof quality control has been achieved: The system is undercontrol.Phase II: Continuous improvement:Continuous improvement is the next step after bringing a processunder control. All aspects of the system remain under constant32scrutiny, by both workers within a system and by management,looking for ways to increase quality and improve productivity.Traditionally, woods workers have not been encouraged tocontribute to the process of improving the system. To besuccessful in achieving continuous improvement, management mustrecognize that the workers have a detailed knowledge of thesystem. They should be the first source of ideas forimprovement. These ideas do not have to be revolutionary butmerely contribute to the process of improvement in some way.Experience gained during the process of implementing acomprehensive quality control system at the Carter Holt HarveyLtd. (Cull) sawmill in Taupo, New Zealand (Collins, 1993), hasshown that it is better to have many small improvements than toattempt a major improvement with one change. At this mill allemployees were involved in the process of improving quality, withsignificant gains in recovery and safety.Continuous improvement can be monitored and observed usingShewhart charts over an extended period of time. As continuousimprovement efforts take effect, this should be reflected in agradual and continuous narrowing of the upper and lower controllimits and lowering of the average level of defects. Common andspecial variation have been discussed above. Phase II isconcerned with reducing the normal operating limits due to commonvariation so that even if special variation causes the processaverage to be shifted beyond the normal operating limits, theproduct specification limits will not be exceeded (Maness, 1993).The product specification limits are the limits set by the33customer, such as a sawmill, and represent the customers desiredlevel of quality. To achieve what is in effect a safety marginbetween the desired output and the actual output, the commonvariation of the system must be reduced until the normaloperating (6 times Standard Deviation) limits of the output ofthe system are narrower than the product specification limits bythe largest possible amount of special variation. The benefit ofreducing common variation to this level, is that even ifsomething goes wrong with the process, and the safety margin isused up, the product will still be within the customersspecification limits.34CHAPTER THREE - PILOT STUDYThe literature review has demonstrated that there is a number oftechniques which can be used to help control the effect ofyarding operations on log recovery value. The next step in thedevelopment of an effective method of controlling losses is toperform a field study of the operating conditions and constraintsoccurring in yarding operations on the coast of British Columbia.This will help to ensure that methods developed for controllingyarding operations will function effectively under actualoperating conditions.In order for a field study to be effective, it should be designedto measure a particular factor. A factor which should be knownfor the development of cost effective quality control proceduresis the long run average loss in recovery value and thevariability in that value. This factor is the basis fordetermining where and how much effort should be spent oncontrolling variation. The difficulties of implementingprocedures for controlling and improving the production processmust be offset by an improvement in the average loss due todamage if there is to be a net benefit to implementing theprocedures. It was decided, therefore, that a pilot study wouldbe conducted to develop and test procedures for determiningaverage value loss under several different commonly used yardingsystems. This information could be used to evaluate furthermeasures to implement production control procedures. Cost35estimates of proposed control procedures were beyond the scope ofthis study.Data collected was used to calculate base line data for thesystems used in the pilot study. This demonstrated theeffectiveness of the methods proposed for data collection andanalysis. The problems encountered were evaluated and solutionsdeveloped or suggested where relevant to the development ofquality control methods for yarding.Methods:A site for the pilot study was made available through cooperationwith the Forest Engineering Institute of Canada (Western) (FERIC)and Fletcher Challenge Canada Limited (FCC). The site wastypical of a setting in a low elevation Cedar-Hemlock-Balsamstand on the west coast of Vancouver Island. The trees in thesetting were felled and bucked prior to any measurements. Threetypes of yarding systems were used in the setting, providing theopportunity to develop base line data for three systems. Table 3shows the relevant details of the site characteristics and Table4 shows the details of the yarding systems for each area. TheAppendix contains location and setting maps.36Site characteristics.Units SKYLINE GRAPPLE HIGHLEADAPEA Hectares 3.5 4.7 7.1SLOPE- Percent 10 15 20averageSLOPE-range Percent 0 - 20 5 - 20 0 - 60TERRAIN GENTLY UNBROKEN RIDGED /ROUGHNESS ROLLING SLOPE GULLIEDESTIMATED Cubic Metres 2,900 3,900 5,900VOLUMEESTIMATED $ 306,000 411,000 621,000VALUEThe volume estimates were based on the average volume per hectarefor the opening obtained from the operational cruise of the area.The value estimates were based on the volume estimates and theaverage market value per cubic metre of the logs sampled for eacharea (Average domestic log selling prices for major coastalloggers; for the month ended December 15, 1992).Table 4 shows the yarder type, cable system and othercharacteristics of the yarding systems.Table 3.37Table 4. Yardinq system characteristics.SKYLINE GRAPPLE HIGHLEADYARDER TYPE Swing Yarder Swing Yarder Mobile SparCypress Model Cypress Model Madill Model 9017280 7280CARRIAGE Butt Rigging Grapple Butt RiggingTYPE W/ chokers & & Rider Block WI chokers &Rider block Rider BlockTAILHOLD Backspar tree Excavator! Stump Excavator! StumpCABLE SYSTEM Running Skyline Running Skyline Running SkylineEach area is labeled according to the common term for the type ofsystem used in that area; skyline, grapple or highlead. Allthree systems are technically running skyline systems because ofthe addition of the rider block to increase lift in each case.The major differences between the three systems were the twomeans of connecting the logs to the system (chokers and grapple)as used by the Skyline and Grapple systems and the two yardertypes (swing yarder and mobile spar) as used by the Skyline andHighlead systems.Data collection:To determine value loss it was necessary to measure and grade anumber of logs prior to yarding and then examine them afteryarding for any damage. Value change would then be determinedbased on the change in volume and grade of the log usingpublished open market log values.The basis for selection of the logs was to provide an adequatesample of all species of logs on the site. As yarding distancecould be a factor in the amount of damage, samples were taken38uniformly over the range of yarding distances to minimize theeffect of yarding distance on the results. Particular attentionwas given to selecting lumber and shingle grades of logs as theywere deemed more likely to suffer significant value loss thanpulp grade logs. It was attempted to select different species inapproximately the same proportions as in the original stand.Logs were selected at random using the guidelines stated below.A target of 200 logs for each area was set, however due tochanges in the system area boundaries, this target was notreached in the skyline area.The criteria for selection of logs in the field were:• The log was not rotten or damaged prior to yarding asindicated by examination of the visible surface of thelog. Rotten logs were excluded because of the difficultyin estimating the deduction for rot required on aconsistent basis.• The log was bucked at both ends. Selecting logs whichwere already broken would create bias in the results.This is because the potential for loss with these logswould be less than with logs which were intact whenmeasured. Large cedar logs which were split or partiallysplit from falling were also excluded for the samereason.• Enough of both ends of the log was visible to scale andapply identification numbers and tags. Where it wasuncertain whether a log could be scaled, it was selected39and left up to the scaler’s discretion to decide whetherit could be scaled or not. Some errors in the fieldscale were found when the logs were examined at thelanding which were adjusted by rescaling and correctingthe field scale.In order to select the target number of logs in each area, it wasnecessary to select as many of the logs which met the abovecriteria as possible. A sampling plan such as a line transect orvariable size plot methods was not used because it would probablyhave resulted in reducing the number of logs selected below thetarget levels. Table 5 shows the details of the logs sampled foreach area:Table 5. Sample statistics.Units SKYLINE GRAPPLE HIGHLEADNUMBER pcs 151 269 274VOLUME 816.3 1337.8 1119.4VALUE $ 109,000 170,000 134,000AVERAGE $/m3 133.73 132.58 119.99VALUE /PCAVERAGE m3 5.40 4.80 4.10SIZE/PCMINIMUM m3 1.03 0.51 0.35VOLUME /PCMAXIMUM m3 25.08 23.84 21.26VOLUME /PCA licensed scaler was supplied by FCC to measure and grade thelogs in order to achieve consistency and accuracy of40measurements. The scale was based on the gross measurements andvisible physical characteristics of the log. No deduction wasmade for hidden defects, in order to minimize inconsistencies inscaling.Data recorded by the scaler at the felling site included:• Log number• Species•Statutory (Ministry of Forests) grade•Log length to nearest 0.2 metre•Top and bottom diameters in Rads (radius in centimetres)The landing scale was preceded by a visual inspection of the log.Only logs which had visible damage or appeared to have beenincorrectly scaled in the field were rescaled at the landing.This substantially reduced the amount of scaling time involvedwith no loss of information because only a change in volume orgrade was of interest in the study.To facilitate identification of logs at the landing, all logswere numbered in a manner that would withstand the roughtreatment the logs would receive in the yarding process. Anumbered aluminum tag was fastened to the butt of the log with analuminum nail. If two logs were so close together that the buttof the selected log was not accessible, a bevel was cut in theend of the log, deep enough that the tag could be fastened andwould be visible when the log was yarded. Logs were alsonumbered using log marking paint and lumber crayon. It was41important that either the marks or tags be visible from the endto identify logs in piles or truck loads.To keep track of all logs that were examined at the landing,records were kept of the numbers of marked logs on each loadedtruck along with the load ticket number, truck number, date andcomments concerning any damage to the log. These commentsprovided the basis for the analysis of damage by type.Data analysis:The scale data was received from the scaler on a computer diskand imported into a spreadsheet file. The landing scale data andother damage related information were entered manually into thespreadsheet and matched with the original scale data. Separatelistings and summaries were prepared for the three differentharvesting systems. Logs which were yarded by a system differentthan originally planned were transferred to the appropriatesample listing. For logs which were not scaled at the landing,the field scale data was used for the landing scale, to makecomparisons possible on a total basis for each system. Any logsfor which scale data was obviously mismatched indicating an errorin observing log numbers, scaling, or in recording data wereeliminated from the sample.Log volumes were compiled using Smalian’s formula. As a check onthe accuracy of the compilation, the volumes calculated for eachlog were compared to the volumes which had been calculated by thescaler using the same data.42The volumes and grades based on the field scale and the landingscale were compared on a log by log basis and any differenceswere attributed to the effects of yarding. Logs which were notexamined at the landing were excluded from the population ofsampled logs.The value loss associated with each damaged log was estimatedusing published December, 1992 Vancouver Log Market values foreach species and grade category according to the British Columbiastatutory grading system. These values were consideredrepresentative of open market log values over the period ofstudy. The purpose of using the Open Market log values was toweight the loss for the higher grade logs. Because only therelative change in value was being measured, no deduction wasmade for hauling, sorting and other costs associated with gettingthe logs to market.Loss in market value was determined for each system. The losseswere calculated in percent of total value yarded to provide afair means of comparison.Results:The results of the pilot study are shown below. These are: anestimate of the average loss for the yarding systems used in thestudy, and a summary of the problems encountered during the studyand how they would be dealt with in the recommended qualitycontrol method.43Average value loss:The average value losses for each system and in total as found bythe study are shown in Table 6. Total damage includes value lossdue to both volume loss and grade change. These loss factorswere grouped together because under the grading rules used, gradeis partially dependent on the same physical dimensions thatdetermine volume.Table 6. Loss in recovery value by system.SKYLINE GRAPPLE HIGHLEAD TOTALLOGS DAMAGED 3 7 6 16(Pieces)TOTAL VALUE LOSS 1,253 1,351 2,176 4,781($)AVERAGE VALUE LOSS 1.1 0.8 1.6 1.2(% OF SAMPLE $)Table 6 shows that the average value loss was 1.2% with lossesranging in value from 0.8% for the grapple system to 1.6% for thehighlead system. These results are within the range of resultsexpected from the literature review. The number of logs damagedin each system was very small relative to the number of logssampled. Because of this, it was concluded that analysis of thetypes of damage would be more appropriate than a statistical testto determine if there was any difference between the results foreach system. This approach is consistent with the objective ofthe pilot study which was mainly the development and testing ofmethods. The conclusion above indicates that a much largersample would be required to ensure that enough damaged logs occur44in the sample to perform statistical comparisons of the results.For this reason, it was decided to investigate methods whichwould not require such extensive sampling for the amount ofinformation gained, such as checksheets and control charts forattributes. Examples of how the results would be monitored usingcontrol charts for attributes and how detailed analysis of thedamage would be done are shown in the following sections.Recommended procedures for using checksheets and control chartsare discussed in chapter 4.The above results demonstrated that average value loss can beestimated using the methods developed for the pilot study.However, because this was a pilot study, the methods used and theproblems found are of more importance than the results. In viewof the number and the nature of the problems encountered in thestudy, the results should be used with extreme caution. Theproblems are discussed below.Monitoring log damage:The damage to logs resulting from yarding over the periods inwhich logs were scaled is shown in Table 7 and plotted on acontrol chart in Figure 6. Although the periods roughlycorrespond to a week, this is not entirely so, because in thethird week of yarding, the type of yarding system changed fromskyline to grapple. Because this is a change in the rationalsubgrouping, the sample for this week was split into two periods:45Table 7. Yardinq results and statistics by period.Scaling of marked logs began in Period 1 (June 18), withbeing grapple yarded to the roadside, scaled, and loaded outimmediately. Then, beginning in Period 2 (July 1), logs whichhad been previously yarded and windrowed with the swing yarderand chokers, were scaled concurrently with loading. In Period 4(July 9), loading and scaling of the main portion of the grappleyarded logs which had also been yarded previously and windrowed,began. Following a shut down from August 1 to November 11,highlead yarding began in Period 7 (November 12). Logs werePERIOD LOGS LOGS PROPORTION UPPER YARDINGSCALED DAMAGED DAMAGED CONTROL SYSTEMLIMIT1 20 0 0.000 0.125 Grapple2 77 0 0.000 0.075 Skyline3 87 3 0.034 0.072 Skyline4 40 2 0.050 0.095 Grapple5 140 4 0.029 0.062 Grapple6 50 1 0.020 0.087 Grapple7 30 0 0.000 0.106 Highlead8 60 3 0.050 0.082 Highlead9 70 0 0.000 0.077 Highlead10 63 1 0.016 0.080 Bighlead11 15 1 0.067 0.140 Highlead12 8 0 0.000 0.183 Highlead13 21 1 0.048 0.122 Highlead14 5 0 0.000 0.226 Highleadlogs46yarded, scaled and loaded out the same day, which continued untilPeriod 14 (February 25).Figure 6. Control chart for log damage (study data).PROPORTION DAMAGED BY PERIOD0.25Z0.10.05 AOu—i____ ____PERIOD—0-—— PROPORTION -—0-— UPPER CONTROL • MEANDAMAGED LIMITFigure 6 shows that the proportion damaged did not exceed theupper control limit during any period. The upper control limitvaries greatly between periods due to the large range of logsscaled each period. This situation is very likely to occur underactual operating conditions, because combinations of variousfactors can cause daily or weekly production to fluctuatesignificantly. A problem which arises because of the shift inthe upper control limit is that shifts in the average may not bewhat they appear to be at first glance. For example, theproportion damaged in Period 8 (0.50) appears lower on the chartthan the proportion damaged in Period 11 (0.67). However,because the upper control limit increases from 0.082 in Period 8to 0.140 in Period 11, the proportion damaged in Period 8 is47actually closer to being out of control than in Period 11.Duncan (1965) discusses this problem, caused by large changes insample size. He suggests the preparation of a stabilized chartin such cases, where the values being recorded are expressed inunits of the standard deviation, instead of proportions. Figure7 shows how the scaled logs for the fourteen periods would appearin a stabilized chart.Figure 7. Stabilized control chart for log damage.PROPORTION DAMAGED BY PERIOD IN STD. DEV. UNITS3 —f , , , , , ¶—3PERIOD• PROPORTION OCENTRE • UPPER —DLOWERDAMAGED LINE CONTROL CONTROLLIMIT LIMITIn Figure 7, the proportion damaged in Period 8 is now closer tothe upper control limit than in Period 11. Stabilized chartsshould be prepared in addition to standard control charts beforeanalyzing the patterns of recorded points when sample sizes varygreatly. Standard control charts should be used in anydiscussions with people not familiar with statistical concepts,because of the increased complexity of the concepts involved inthe stabilized chart.48With stabilized charts it is also possible to place warninglimits at one and two sigma(o) levels without significantadditional calculation. If warning limits are set at the onesigma level, in this case, Periods 4, 8 and 11 would be outsidethe warning limit and would be investigated more closely.The pattern of the points in Figure 7 does not meet any of thecriteria for judging out of control situations as discussedabove. Therefore it can be concluded that the changes from oneyarding system to another that occurred during the study did nothave an effect which could be distinguished from randomvariation. Because the effect of the changes, if any, cannot bedistinguished from random variation, the changes cannot beclassified as special or assignable causes of variation, andtherefore, any variations in the results from sample to samplecannot be attributed to the changes in the yarding system.Analysis of Damage:The frequency of damage by different types is shown in thehistogram in Figure 8. This chart shows that broken tops andslabbing are the most frequently occurring types of damage.49Figure 8. Histogram of damage by type.The Pareto chart in Figure 9 shows the relative value losses fromdifferent types of damage.Figure 9. Pareto chart of damage by type.VALUE LOST ( $) • CUMULATIVE VALUE LOST(%)5FREQUENCY OF DAMAGE TO LOGS BY TYPE OF DAMAGE31Cl)00U120BROKEN TOP BROKEN END SHATTERED SLABBEDENDDAMAGE TYPETOP BROKEN BUTT SPLITSLABBED BUTTDAMAGE BY TYPEE-’Cl)050004500400035003000250020001500100050001787.100 1009080706050403020100c’PHU)I-BROKEN END BUTT SPLIT690353I ISLABBEDDAMAGE TYPEOTHER50From this Figure, it is evident that the largest value lossoccurs as a result of broken ends, $1,951, with the next largestloss arising from butt splits, $1,757. Together, these two typesof losses account for nearly 80% of the value lost. This findingis consistent with the Pareto Principle as discussed above. Theconclusion from this analysis is that efforts to reduce damageloss by reducing broken ends will provide the greatest gains invalue recovery.Procedural problems:The major problems encountered in applying the proceduresdeveloped for the study were:Number of logs to be marked:It was recognized at the start of the study that accuratemeasurement of losses during the yarding phase could only beachieved by measuring logs both before and after yarding. Thepre-yarding dimensions of broken logs could not be estimatedafter yarding by measuring their component parts, mainly due tothe problem of locating and identifying the broken bits which getleft in the field. Obviously, it was not possible to tell inadvance which logs would be broken, nor in which areas thebreakage would occur. Also, only a small percentage of logs wasexpected to be broken, though not as few as actually occurred.These factors meant that a large number of logs spread evenlyover the study area had to be measured in order to ensure that arepresentative sample of damaged logs would be obtained.51Recording and analysing the data was very time consuming with somany logs sampled. The recommended quality control system isbased on detailed recording of circumstances associated with thelogs which are actually damaged, and should therefore reduce theamount of recording and analysis time.Productivity:Productivity for the marking phase of the study was reduced bythe difficulty in locating sufficient logs which met theselection criteria. Slash, other logs and heavy brush were thelargest impediments to movement on the ground and selection oflogs. The conditions were considered normal for recently felledstands on the west coast of British Columbia. Actualproductivity of marking was from 20 to 40 logs per day by a two-person crew depending on the brush conditions and the conditionof the logs. About 750 logs were selected, requiring 25 workingdays to tag and mark. The total time taken for marking, althoughnecessary for obtaining baseline data, would not be acceptableunder operational conditions, and therefore further supports therecommended quality control system.Scaling difficulties:It was difficult to see enough of the log in some cases tomeasure and grade it accurately. Where the scaler judged that hecould estimate the measurements with reasonable accuracy, thiswas accepted. The alternative to accepting some reduction in thelevel of accuracy would be to drastically reduce the number of52logs available for selection and reduce the effectiveness of thestudy. Under current coastal operating conditions, this will bea common occurrence, and it may be difficult to overcome thisproblem unless alternate methods are developed not requiringscaling in the field.Coordination with operations:One of the problems with scaling at the landing was being able toexamine and scale the selected logs while yarding and loadingoperations were in progress. The main concern was the safety ofthe scaler and study personnel. It was necessary to carry outthe examination and scaling of the logs in the area where theloader, and sometimes the yarding machine, were operating. Logswere constantly being moved by both machines. The problem wasresolved in the pilot study through coordination with the loaderoperator. The operator scanned the logs as they were being swungout of the way of the yarder or removed from the windrow pile andsignaled the scaler if any logs had markings on them. The loaderoperator also swung the logs to one side if it was necessary forthe scaler to examine the log more closely. For longer-termstudies of value losses and successful implementation of aquality control system, procedures need to be developed which donot interfere with operations.Identifying logs:A major problem anticipated at the beginning of the study was thedifficulty of identifying selected logs in a way which would53withstand the forces of yarding and still be readily visible froma distance and at the landing. Different types of markingsystems were considered, but the best solution appeared to be acombination of all three types available: tags, log markingpaint, and crayon. Improvements are still needed in the methodsof marking and identifying logs as indicated by the fact that0.5% of logs were not accounted for in the Highlead area, thelast area to be marked and yarded.Sensitivity of scaling to log quality factors:The statutory grading system is dependent on physical dimensionsand some of the quality characteristics of the log. Qualitycharacteristics which can effect log value, such as bark removal,scraping, and gouging, and other types of damage which may happento logs during yarding operations, are ignored if the effect ongrade or volume is not significant. Logs which did not showsufficient damage to change these factors under the currentgrading system were not rescaled, although they might have haddamage of the above types which would result in a significantdrop in recovery value at the sawmill. Producers in othercountries have developed elaborate grading systems whichrecognize the needs of different markets. Log grading systemsfor coastal operations need to be developed based on end useproducts in order to maximize the recovery value of logs andminimize waste. For the purposes of determining base line dataon value loss, use of a sawing simulator to estimate the volumeof various grades of products which could be produced from the54selected logs should be investigated. This is beyond the scopeof this thesis.Bias in log selection:Logs were selected according to the criteria given above.Although a large number of logs was selected, the selection wasnot entirely representative for two reasons:Low value pulp logs were selected relatively less frequently thanhigh value lumber grade logs. The basis for this was that thepotential value loss for pulp grade logs would be small becausetheir value was already very low. Pulp logs can suffer a valueloss because, even though they are in the lowest statutorygrades, there is still a narrow range of values for them. Theemphasis of the pilot study was on testing methods of measuringthe loss in log value resulting from log damage. Therefore, theconsiderable extra effort required to obtain a representativesample of all log sizes was not considered to be worthwhile.This policy would undoubtedly cause some undetermined bias in theresults.Another source of bias results from the fact that logs which wereburied under piles were less likely to be included in the samplebecause of the difficulty of measuring and scaling them. Thiswould have an effect on the average value loss because logs whichwere buried under piles of other logs would be more likely to bebroken when the logs are pulled out of the pile by the yarder.The amount of bias caused by this omission could be significant.55Without further study, the amount of the bias is impossible toestimate.Unexamined logs:A number of logs were selected which were not examined at thelanding. There are two main reasons why these logs were notaccounted for: they reached the landing in one or more pieceswhich were not identified because the tags or other marking weremissing or not visible, or they were broken up during yarding andno identifiable pieces reached the landing.If the logs had simply been missed at the landing, but were notdamaged, the measured average loss would be unbiased becausemissing logs were excluded from the sample. On the other hand,if the logs had actually been broken in yarding and notrecovered, the average value loss would be understated because noloss was recorded for missing logs. How many of the missing logswere actually broken is not known. A residue survey of thesetting was done by FERIC after yarding, and no marked logs wereobserved, which would tend to support the conclusion that most ofthe missing logs actually reached the landing but were notidentified. Further study should be done on the problem ofadequately marking the logs when collecting base line data andamage losses. The recommended quality control method reducesthis problem by making notes on log damage as the logs are beingyarded, thereby making it unnecessary to mark logs.56Recovery improvement information:Data necessary to calculate value loss was collected on each logthat was inspected at the landing. However, only minimalinformation related to the type of damage or the causes wasrecorded. This was due mainly to the constraints imposed by thestudy design and the operating conditions. Difficulties inidentifying logs and observing and recording events during theyarding process stemmed partly from the large number of logswhich were selected. More detailed information on the type ofdamage and the causes would be useful in efforts to improvemethods and reduce value loss. The recommended quality controlsystem is designed to provide this information.Value and grade relationship:The estimate of value loss would be considerably more useful formaking management decisions if a set of values had been availablewhich reflected the actual value of the various grades ofproducts which could be produced from the logs. However, thesetypes of values are not available for logs from old-growthforests, on the coast of British Columbia, on a current basis, inspite of the high values that are involved and the significanteffect that mill recovery can have on log recovery value. Atpresent the only values that are available are based on thestatutory grading system which is based primarily on treatinglogs as a commodity product. Statutory grades are based ongeneralized physical characteristics as defined in a set of ruleswhich do not change with market conditions. The market value of57a log based on the statutory grades may vary substantially fromthe actual value of the log to an individual converting mill. Asa result, estimates of average loss based on these values may bemisleading.An approach to this problem which should be studied further, isto use a sawing simulator to determine the value of logs, beforeand after yarding, based on the potential recovery of variousproducts from the logs. It would be necessary to obtain currentmarket values which could be related to the grades determined bythe simulator. It would probably be easier to do this and letthe simulator carry out the conversion of the logs, rather thanto try to develop a system of log values based on the grade of awhole log. This would also require very detailed assessment oflog dimensions and condition. Development of this approachshould also help to resolve problems of scaling logs as discussedabove.58CHAPTER FOUR - RECOMMENDED QUALITY CONTROL METHOD:The primary objective of this thesis was the development of aquality control system to control losses in log recovery value inyarding operations. The method must be practical and usableunder operational conditions. The literature review has shownhow control charts and other statistical tools can be used tocontrol value loss. The pilot study has shown some of theproblems which have to be considered in the development of themethods. The following quality control procedures arerecommended as a means of meeting the objectives.Operational quality control procedures:The procedures recommended for controlling log damage underoperational conditions consist of two major components:• a checksheet for gathering information in the field onthe prevailing conditions, the types of damage, and thepossible causes of damage which occurs during yarding,and• a control chart for displaying over time, the pattern ofvariation of the proportion of logs damaged as determinedfrom the information recorded on the checksheet.The application of each of these components is discussed below.Checksheet for recording damage information:A checksheet is the most suitable means of collecting the data.The form should allow for description of the operating conditions02 0) 020)0)0 02 (I)0QG)02I.z 0) (2 002 0 (2 C/)0) C) 0 0 ‘1 02C) H 0 0C. 0 HCi)020 02 Cn x0 H 02 020) C) 0 02 C) 002 0 C/)0 H C) 0 H H Q002020202WWQt2 z 020 ‘---C/)H2 Ci) .,--I’,: 1 i/) !z02a‘0hj11j‘Q0QI-F-I-IctLQI-hCDt11II1CDI-ICDCDPI-ICDCDCDI- o0I-’ct)f•••CDC)CDCDC)0QCDU)C)CDI-U)U)C)CDC)CDCDCDC)HiCDC)0U)(-I-‘-IJU)H5CDU)0CDI-’IICDCtU)CDd,<I—i0p,I-II-ICDI-•I-’QU)CDF’•I-’r1jlQi.QHCDCDCD0QMQ‘tPCD0I-hH5U)U)Q)0F-U).QIIi.QI-i.CDCtCtCDI-•)CDQCDU)F-aU)U)U)I-CD)0CDU)I.eP)U)P)ctU)II0CD<CD0U)U)dU)I-’•H•CDU)MH5•CD0I-J..QI-IU)CDCtU)C)CD009)H5F-9) CD—IUiH 0 HH-HH.HH‘30H 060Periodically, for example, weekly, a number of logs yarded in agiven setting are sampled by observing them being yarded andrecording any damage that occurs to them. As discussed above,the most important factor to be considered when determining thesample size and frequency of sampling is the rational subgroup.The recommended sampling method is to observe a full days yardingwith each sample, to obtain as large a sample size as possible,however, each sample should coincide with a period of similaroperating conditions. Therefore, if a change occurs in theyarding conditions during the day, such as a change in terrain,or a change in the yarding machine or operator which could havean effect on the proportion of logs damaged, a new sample shouldbe started.The frequency of sampling depends on the level of damagecurrently being experienced. If the level is high, the frequencyof observations will have to be increased, to perhaps daily.Once the frequency of damage has been reduced to acceptablelevels, the frequency of observations can also be reduced. Thedetermination of acceptable levels will depend mainly on cost andbenefits of sampling as discussed above.Actual observation of damage will begin when the log startsmoving toward the landing and ends when it reaches the landing.The observer should be in a position to observe the logsthroughout this operation. The observer will need to be in radiocontact with the yarding machine operator, to enable him torequest that specific logs be set aside at the landing for61inspection. If a log is damaged and only part of the log reachesthe landing, the observer will estimate the proportion of the logwhich is lost. Exact measurement of the lost portion is notrequired, as the value does not need to be calculated.Many different types of damage can occur more or less frequently,and the observer will need to make careful notes, includingdiagrams, if necessary, to fully document the damage. Thecriteria for recording damage should be to record any type ofdamage which affects the value of the log or the potentialproducts which can be produced from the log. Judgment will berequired, particularly in the early stages of the implementationof the quality control system. Consultation between the observerand a prospective end user of the log, such as a sawmill or logtrader, should take place to resolve questions as to the effectsof various types of damage on recovery value.Observations during the yarding process combined with inspectionat the landing in accordance with the above guidelines should besufficient to determine the frequency, type, and severity ofdamage. After the system has operated for a while, the methodsshould be evaluated to determine if they are operatingeffectively and changes made as necessary in the criteria forassessing damage.Once the quality control system has been implemented on aperiodic basis and the problems in data recording that arise havebeen dealt with, consideration can be given to having thechecksheet maintained on a continuous basis by a member of the62yarding crew such as the engineer or hooktender. Although thiswould require some investment in training to achieve this goal,it is an ideal arrangement because the crew would then beresponsible for monitoring its own performance. This shouldprovide more relevant information and reduce the overall staffingrequired to collect the data. However, it is not likely thatthis would be feasible without improvements in current industrylabour-management relations. The checksheet should be used fromthe beginning as the basis for informal discussions between crew-members and supervisors on quality and damage control. Becauseyarding crews consisting of two to six workers are all involvedin the process of yarding as a team, any control procedure shouldbe applied to the crew as a group or team, rather than asindividuals. Discussions of quality considerations and actualparticipation in the control process, will help to improve theeffectiveness of the system.Control chart for monitoring log damage:A control chart for attributes would be used to monitor logdamage. The control chart would show the pattern of log damageover an extended period of time using information from thechecksheets.An example of a control chart for attributes designed for thispurpose using sample data is shown in Figure 11.63Figure 11. Control chart for yarding damage.PROCESS CONTROL CHART DIVISION SETTING NO. PERIODI LOGS - - MEAN ——k—-UPPER —c’—— LOWERDAMAGED CONTROL CONTROLLIMIT LIMIT0.0600 ---0.0500 - -- - - - - - - -O %. —L I0 ,.I - -. — — -4\ I 4 - —/..‘/— % — — —/ ——/00.0400---’ —/I I I I I I I I I• • —0 I I I I I I0 0200 - - - -.ç- -- --içi -- - -0.0100--- —o.oooor..1 2 3 4 5 6 7 8 9 10 11 12 —SAMPLE NUMBERSAMPLENO. 1 2 3 4 5 6 7 8 9 10 11 12TOTALSAMPLE SIZE 550 249 301 400 275 255 600 175 244 255 350 277 3931DAMAGED 8 3 8 17 5 2 0 3 6 13 6 5 76DATE /TIMEOBSERVERThe example provides room for up to 12 twelve samples but can bemodified to accommodate more samples to provide a longer termview if required. The calculation of the control limits andplotting of the data were covered in the literature review. Theexample shows how the control limits would vary with a varyingsample size. In this example, sample sizes were relativelyconsistent, therefore it was not necessary to prepare a64stabilized control chart as discussed above. The advantage tousing proportions in the chart rather than standard deviationunits is that the chart will be easier to understand by peoplewho are not familiar with statistics. If large variations insample size occur, stabilized charts should be prepared inaddition to standard charts, and the results discussed only withmanagement and others who are familiar with statistics.The control limits calculated initially should be consideredpreliminary limits. As a detailed history of log damage iscreated through successive samples, the normal operating patternshould become evident. The length of the initial period willdepend on how long it takes for everybody involved in the datacollection process to become familiar with the requirements, andfor necessary modifications to be made to the checksheets andcontrol charts. Establishing control of the process will resultin a improved long term process average.Long term application of control charts:When the initial period has passed, the control limits shouldstill be recalculated periodically, for example monthly, becauseof long term reductions in the process average resulting from theimplementation of the quality control program. Long term controllimits should be calculated using the moving average method, witha sliding period such as two to three months, which shouldprovide enough time to eliminate short term fluctuations. Longterm control limits may also be adjusted to reflect theobjectives of the chart. For example, lowering the upper control65limit would cause more effort to be spent on investigatingspecial causes of variation. On the other hand, if managementdecides that current performance levels are acceptable and thatproblem investigation efforts should be concentrated in otheroperating areas, such as falling and bucking, the control limitscould be raised. Only very seriously out of control situationswould then be investigated.Charts should be distributed as soon as they are prepared, whichshould be on a regular basis, to both the workers and thesupervisors who are involved in the operations being monitored.The charts can be discussed in informal meetings with the crew,along with the checksheets as mentioned above. As people becomefamiliar with the charts and checksheets, the workers who arebeing monitored by the charts can become more involved in thepreparation of the charts. For example, once long term controllimits have been established, they can be put on the charts inadvance and the only thing that needs to be plotted is thepercentage of logs damaged in each successive period, which couldbe done by one of the crew-members.The timeliness of the review of the charts with the crew is veryimportant. Except under very unusual conditions, there is noreason why the charts cannot be prepared in the field and shownto the yarding crew which has just been observed, preferably atthe end of the same day on which the observations are made sothat the details of the conditions will still be fresh in their66minds. Comparing the recorded data with their impressions shouldprovide valuable insight into causes of variation.A benefit of keeping continuous records of log damage andbreakage performance on a setting by setting and crew by crewbasis would come from increased awareness of the frequency ofbreakage by all concerned, including the crews. By monitoringthe number of logs damaged using the above methods, a simple andstraight forward but detailed history of performance can bedeveloped. Every member of the crew will thus become consciousof their performance and will see the times when it is better orworse than average with a positive, problem solving attitude.They will also appreciate more fully the need to reduce damageand will be drawn into the process of improving performance.This sense of participation is likely to be one of the mosteffective tools in improving performance and value recovery.Analyzing the system:Histograms and Pareto charts can be used to further analyze datagathered and assist in eliminating special causes of variation.Examples of these types of statistical tools have been discussedin the literature review. The uses of analysis tools and theprocesses of problem solving are complementary to control charts.However, further discussion is beyond the scope of this thesis.67CHAPTER FIVE - DISCUSSION:Literature review:The literature review has shown that control charts forattributes can be used where measurement of a variable isimpractical. Damage to a log is an attribute which can bedetermined by visual inspection. Control can be established bymonitoring individual log damage and providing feedback toworkers and supervisors on the results of operations.In the discussion of the relationship between sample size andsensitivity of the sample, it was shown with the use of operatingcharacteristic curves, that larger sample sizes are better atdetecting changes in the process average. Sample sizes wereestimated using methods suggested by Juran and Duncan, both ofwhich indicate that a sample of approximately 440 logs isrequired to detect a shift of 3a with a 95 percent certainty.The discussion also described how the criteria for determiningout of control points in control charts are based on the assumedrandom nature of the variation in a system which is undercontrol.The literature review discussed methods of analysing a systemincluding checksheets. Information leading to development ofways of reducing log recovery value losses can be collected atthe same time that logs are being monitored for damage by using aproperly designed checksheet.68Pilot study:An essential first step in the development of any quality controlprogram is determining the operating characteristics of thesystem. In this case, value loss in yarding is thecharacteristic of concern. It is important, therefore, to havesome knowledge of the actual losses which are occurring. Theinitial tests performed in the pilot study provided some of thisessential knowledge in the form of base—line data for severaldifferent systems used in the yarding, through comparisons of logvalues before and after yarding. The results indicated valuelosses were approximately 1 percent, which is consistent with thefindings in the literature review. However, the tests requiredextensive time selecting, measuring, marking, and observing theyarding of logs, which for the most part were unchanged in theyarding process. The percentage of individual logs damaged wasrelatively low: only sixteen logs out of nearly seven hundred(see Table 6). The amount of work involved in the detailed testsand the disruption of normal operations which unavoidablyoccurred, make it impractical to apply the pilot study methods onan operational basis. Because of the low levels of breakagefound, however, it would be practical to focus attention onoccurrences of individual log damage.The results of the pilot study emphasized the need for a means ofcompensating for the effects of widely varying sample sizes. Itwas shown how a stabilized control chart would eliminate thisproblem.69Recommended quality control system:These procedures involve the use of control charts as discussedin the literature review above. Control charts are a methodwhich has been found to be effective in many industries andsituations. Other quality control methods for analyzing a systemwere also discussed in the introduction above and should be usedas appropriate.The procedures suggested for controlling log damage underoperational conditions consist of two major components:• A checksheet for gathering information related to damageduring yarding, and• a control chart for attributes to summarize and monitorthe log damage recorded on the checksheet.These procedures are recommended as a means of overcoming two ofthe main problems identified in the pilot study; collectingsufficient data to begin solving the specific problems involvedin improving value recovery, and monitoring and controllingoperations on a current basis without costly and time consumingmeasurements of large numbers of logs. Checksheets are aneffective means of collecting this information by providing spacefor recording large amounts of information in a systematicmanner. Control charts for attributes are suited to the normalsituation in harvesting operations in which the numbers of logswhich are actually damaged are relatively small. Eliminatingmeasurements of logs makes it feasible to examine the large70number of logs which is required to ensure that samples include arepresentative number of damaged logs.Using attributes control charts also provides a means of controlin the interim period while the development of methods of scalingand grading logs which reflect actual value loss potential isproceeding. When improved methods of scaling and grading aredeveloped, they can be tested and incorporated into controlsystems to monitor actual value loss.Other advantages of using control charts for attributes in therecommended manner are: Sample bias resulting from problems withmarking and sampling will be eliminated. This is because alllogs yarded during the observation period will be part of thesample. Unexamined logs will not be a problem with therecommended method. It will not be necessary to mark logs inadvance of yarding, and therefore, problems with identifying andaccounting for logs at the landing will be eliminated.Conclusions:In this thesis, quality control methods have been examined andfield trials have been performed, with the objective ofdeveloping a quality control method to improve value recovery inyarding operations. The following conclusions have been derived:Control charts can be used to improve value recoverythrough monitoring of operations, assisting in theanalysis of special causes of variation and providingfeedback of results to workers and management.71• Estimates of value loss are approximately one percent foroperations using current yarding systems in old growthstands on the west coast of Vancouver Island. This lossis consistent with losses estimated by others.Differences in results between systems could not bedistinguished from random variation.• The recommended approach for yarding operations is:Establish control using charts for attributes and checksheets to remove special causes of variation, thencontinue improvement of operations by reducing commonvariation.• Savings from improving value recovery may be substantial.An improvement of ten percent in the estimated averagevalue loss of one percent would result in an increase inrevenue of $30,000 annually, for the typical woodsoperation discussed above. This should cover the cost ofimplementing the recommended system. Further benefitsshould be realized over time as continued improvement isachieved.• Difficulties in scaling, grading and collecting data oninitial log values need to be resolved. The currentscaling system is designed to measure volume and gradefor statutory purposes and does not measure variations inquality which do not effect statutory value. Gradingrules followed for statutory grading of logs do notreflect the potential value recovery from converting logsto their highest value product and therefore value losses72in higher log value categories are likely to besubstantially understated if based on statutory grades.Methods of data collection need to be developed whichwill make it possible to safely, efficiently, andaccurately measure and grade logs before and afteryarding.73BIBLIOG1APHYAverage domestic log selling prices for major coastal loggers;for the month ended December 15, 1992. 1992. Forest IndustryTrader. Airan Industries Ltd. Vancouver. 3:43. 6 p.Collins, Stuart G. 1993. Unpublished comments on QualityImprovement. Carter Holt Harvey Timber Ltd. Taupo, NewZealand.Dey, Jon. 1993. Quality improvement tools. Proceedings of thejoint LIRO/FIEA Seminar: Quality issues in harvesting andprocessing. June 9-11, 1993. Rotorua, New Zealand. (inpress).Duncan, Acheson J. 1965. Quality control and industrialstatistics. 3rd Ed. Richard D. Irwin, Inc. Homewood,Illinois. 992 p.Gabrielsson, Lars and Tommy Helgesson. 1989. Wood damage - awaste with resources. Skogsarben Resultat no. 24. 1989. 5 p.Gitlow, Howard S. 1987. The Deming guide to quality andcompetitive position. Prentice - Hall, Inc. EnglewoodCliffs, New Jersey. 247 p.Grant, Eugene L. and Leavenworth, Richard 5. 1988. Statisticalquality control. 6th Ed. New York. McGraw Hill Inc. 714 p.Hansen, Bertrand L. 1963. Quality control: theory andapplications. Prentice-hall, Inc. 498 p.Juran, Joseph M. 1962. (ed.). Quality control handbook. 2ndedition. McGraw-Hill Book Company. New York. 1154p._______________and Gryna, Frank M. 1993. Quality planning andanalysis: from product development through use. 3rd edition.New York. McGraw Hill Inc. 634 p.Maness, Thomas C. 1993. Principles of industrial quality controlcourse notes. University of British Columbia. Vancouver,Canada. Unpublished.McIntosh, J. 1968. How close-U harvesting could effect woodvolumes and logging practices. Canadian Forest Industries,88(7): pp. 44—51.McNeel, Joseph F. 1993. Personal communication.Murphy, Glen and Alastair Twaddle. 1985. Techniques for theAssessment and control of Log Value Recovery in the New74Zealand Forest Harvesting Industry. Paper presented at the9th Annual Council on Forest Engineering Meeting. September29-October 2, 1985. Mobile, U.S.A. pp. 43-47.Oakland, John S. and Followell, Roy F. 1990. Statistical processcontrol: a practical guide. 2nd Ed. Oxford. HeinemannNewnes. 431 p.Scherkenbach, William W. 1986. Deming route to quality andproductivity, road maps and roadblocks. Washington, U.S.A.CEEPress Books, George Washington University. 145 p.Stuart, W.B., Grace, L.A. and LeBel, L.G. 1993. Harvesting,Transport, Primary Processing, and Total Quality Management.Proceedings of the IUFRO P.3.07-01 Harvesting and productQuality Inaugural Meeting. June 9-11, 1993. Rotorua, NewZealand (in press).Williston, E.M. 1979. Opportunity areas and leverage points. In:Electronics in the sawmill. Proceedings of the electronicsworkshop. Sawmill and plywood clinic. Portland, Oregon.March,1979. pp. 14—18.75APPENDIX - MAPSMap 1. Location map of Waibran Creek.Map 2. setting map showing areas yarded by each system.76L)CLd6LMAP .WPLBLAI) c466 Nor TO SCALEçoLtpAI‘5Tvby S,TEz-Pc(;7b(1%o.•%,7_I.--‘;“o°0.C”FC’.1kN:ris—0-(Elb‘<Vb•b—n—crC) 3Zr.u0%,JI’) 7’I 0z

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0075161/manifest

Comment

Related Items