Managing lumber stiffness variation in a sawmill : a statistical quality control approach

Tshavhungwe, Vhuhwavho (2019-01)

Thesis (MScFor)--Stellenbosch University, 2019.

Thesis

ENGLISH ABSTRACT: Several studies over the past two decades in South Africa showed that graded structural lumber from certain young pine resources did not conform to the stiffness requirements of SANS 10163- 1(2003). The broad aim of this study was to propose a quality control system to South African sawmills, which will ensure the supply of structural lumber that is safe and reliable in its end-use application. The focus was on modulus of elasticity (MOE), since previous studies showed that this was the property most at risk of non-compliance to current standards. The specific objectives of this investigation were the following: (i) to quantify the amount of MOE variation within and between bundles of graded structural South African pine lumber from two sawmills, (ii) to attempt to define what is acceptable MOE variation for end-users of structural lumber in South Africa, and (iii) to evaluate the current proposed SANS 1783-5-2 quality assurance system and compare it with other possible systems in terms of its efficiency to ensure safe and reliable structural lumber in terms of lumber MOE. Board MOEs from two anonymous sawmills (Sawmill A and B) in South Africa processing pine saw logs were measured for this study. Twelve months of data were available for Sawmill A and 2 months of data for Sawmill B. Only data for 38x114mm lumber from both sawmills graded as either utility grade / XXX (non-conforming), or structural grade S5 were used in the analysis. The dynamic MOE (MOEdyn) for each board was measured using a resonance type stress grading device. The data was sorted into bundles using the intake bundles sizes (190 pieces per bundle). Variation in MOE was analysed and a reliability analysis performed. Various quality control methods were compared in terms of their ability to detect production periods where a large percentage of bundles did not conform to the reliability requirements in terms of both the serviceability and ultimate limit states. Results showed that the mean MOE and 5th percentile MOE values of the full population of S5 graded 38x114 mm pieces were above the required SANS 10163 levels for both Sawmill A and Sawmill B. Over time and between bundles, however, there were fairly large changes in the MOE means and 5th percentile values between the two sawmills. In general, Sawmill A had higher mean bundle MOEs than Sawmill B. Sawmill B on the other hand had much lower variation in MOE and a much smaller range in bundle mean MOE values. In terms of reliability, it was found that overall for all the bundles produced in the respective sampling periods, 1.38% of the bundles did not conform to the required reliability index for the serviceability limit state (152 bundles from a total of 11 046 bundles produced). Less than 0.1% of the bundles did not conform to the required reliability index for the ultimate limit state (8 bundles from a total of 11 046 bundles produced). It was observed that the variability of MOE played a relatively larger role in determining acceptable reliability than the mean MOE. The EWMA chart seem to be the most effective quality control method at detecting production periods where a large percentage of bundles do not conform to reliability requirements. The current proposed quality control system based on a moving average (SANS 1783-5-2), as well as the CUSUM methods also seem to be able to effectively detect out-of-control production periods. However, the stop-production signal threshold in the current proposed quality control system (SANS 1783-5-2) may need to be changed in order to ensure an efficient system. The ARIMA method did not prove to be effective. Although increased sampling frequency enables quicker detection of outof-control MOE, the current proposed sampling frequency of 1 out of 1 000 pieces seems to give acceptable results. Additional research is recommended including a pilot study at several sawmills evaluating the lumber quality control procedure recommended from this study. In this case the static MOE values need to be measured instead of dynamic MOE.

AFRIKAANSE OPSOMMING: Verskeie Suid Afrikaanse studies oor die afgelope twee dekades het getoon dat gegradeerde, strukturele hout vanaf sommige jong, denneboom-bronne nie voldoen aan die styfheidsvereistes van SANS 10163-1 (2003) nie. Die oorkoepelende doelwit van hierdie studie was om ʼn kwaliteitsversekeringstelsel voor te stel vir Suid Afrikaanse saagmeulens wat sal verseker dat veilige en betroubare hout verskaf word aan die eindgebruiker. Die fokus van die studie was op die elastisiteitsmodulus (MOE), aangesien vorige studies gewys het dat dit die eienskap is wat die grootste risiko toon vir nie-nakoming van standaarde. Die doelwitte van die studie was: (i) om die variasie in MOE tussen bondels en binne bondels van gegradeerde Suid Afrikaanse dennehout vanaf twee saagmeulens te kwantifiseer, (ii) om te poog om te definieer wat aanvaarbare MOE variasie is vir eindgebruikers van strukturele hout in Suid Afrika, en (iii) om die huidige voorgestelde SANS 1783-5-2 kwaliteitsversekeringstelsel te evalueer en te vergelyk met ander moontlike stelsels in terme van die vermoë om veilige en betroubare strukturele hout te verseker in terme van MOE. Plank MOE’s vanaf twee anonieme Suid Afrikaanse denne-saagmeulens (Saagmeul A en B) is gemeet vir hierdie studie. Twaalf maande se data was beskikbaar vir Saagmeul A en 2 maande vir Saagmeul B. Slegs data van produksie van 38x114 mm planke wat gradeer is in algemene graad (XXX) of strukturele S5 graad is gebruik. Die dinamiese MOE (MOEdyn) van elke plank is gemeet met behulp van ʼn resonansie-tipe graderingsapparaat. Die data is sorteer in bondels van 190 stukke per bondel. Variasie in MOE is ge-analiseer en ʼn betroubaarheidstudie is uitgevoer. Verskeie kwaliteitsversekeringsmetodes is vergelyk in terme van hulle vermoë om produksieperiodes te identifiseer waar ʼn groot hoeveelheid bondels substandaard was. Resultate het aangedui dat die gemiddelde MOE en 5de persentiel MOE van al die bondels gegradeerde S5 38x114 mm planke hoër was as die vereisde SANS 10163 waardes vir beide Saagmeul A en Saagmeul B. Oor tyd was daar egter redelike groot veranderinge in gemiddelde en 5de persentiel MOE waardes van elke saagmeul asook tussen die twee saagmeulens. Oor die algmeen het Saagmeul A hoër gemiddelde bondel MOE waardes gehad terwyl Saagmeul B laer variasie in MOE gehad het. In terme van betroubaarheid is gevind dat 1.38% van alle bondels oor die studietydperk nie voldoen het aan die vereiste betroubaarheidsindeks vir die diensbaarheidslimietstaat nie (152 bondels uit 11046 bondels). Minder as 0.1% van die bondels het nie voldoen aan die uiteindelike (“ultimate”) limietstaat nie (8 bondels uit 11 046). Daar is waargeneem dat die variasie in MOE ‘n relatiewe groter rol speel in bepaling van aanvaarbare betroubaarheid as gemiddelde MOE.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/105636
This item appears in the following collections: