In the sawmill industry, it can be very profitable to monitor the dimensions of sawn boards so that operators quickly can detect errors and take cor-rective action. In this master’s thesis project, Statistical Process Control (SPC) methods have been implemented to achieve this. SPC is a set of statistical methods whose purpose is to minimize the variations in an in-dustrial process. In particular, the SPC method used here is the control chart, which with an upper and lower control limit quantifies the bounds of natural variation. To find the most suitable control chart, five control charts monitoring the process mean, and two monitoring process variability were tested with help of both a simulation study and an empirical evaluation. The result of the evaluation was that the ”Average Moving Range” chart was regarded the most suitable for changes in process mean, and the Range chart was regarded as the best at detecting changes in process variability. Both charts are constructed for individual boards and not subgroups of boards (as is more common) due to compatibility reasons with the existing measurement practice. The two methods were deemed to be quite able to detect process changes, but some results indicate that the methods might work better for double arbour saw lines than single arbour ones. / Det kan vara mycket lönsamt för sågverk att övervaka mått på plankor så att personal snabbt kan hitta och åtgärda fel som uppstår i processen. I det syftet har det här masterarbetet gått ut på att implementera statistisk processkontroll (SPC) för råmåttkontroll på sågverk. SPC är en mängd olika statistiska metoder vars syfte är att minimera spridningen i en tillverkningsprocess. Den metod som är i speciellt focus i det här arbetet är det så kallade styrdiagrammet som med en övre och undre gräns kvantifierar hur stor den naturligt förekommande spridningen är. För att finna det mest lämpade styrdiagrammet utvärderades fem styrdiagram som övervakar processens medelvärde och två styrdiagram som övervakar processens spridning. Denna utvärdering bestod både av en simuleringsstudie och tester gjorda för empiriskt data. Utvärderingen resulterade i att det så kallade ”Average Moving Range” diagrammet rekommenderades för övervakning av medelvärde och ett räckviddsstyrdiagram rekommenderades för spridningen. Båda styrdiagrammen konstruerades för enskilda plankor och inte för stickprov av flera plankor (vilket är vanligare) på grund av kompatibelitetsskäl med gängse mätmetodik. De båda metoderna ansågs vara ganska bra på att upptäcka processförändringar men vissa resultat tyder på att metoderna kanske fungerar bättre för sågverk med mötande klingor än enaxliga sågverk.

Bayesian process control is a statistical process control (SPC) scheme that uses the posterior state probabilities as the control statistic. The key issue is to decide when to restore the process based on real-time observations. Such problems have been extensively studied in the framework of partially observable Markov decision processes (POMDP), with particular emphasis on the structure of optimal control policy. Almost all existing structural results on the optimal policies are limited to the two-state processes, where the class of control-limit policy is optimal. However, the two-state model is a gross simplification, as real production processes almost always involve multiple states. For example, a machine in the production system often has multiple failure modes differing in their effects; the deterioration process can often be divided into multiple stages with different degradation levels; the condition of a complex multi-unit system also requires a multi-state representation. We investigate the optimal control policies for multi-state processes with fixed sampling scheme, in which information about the process is represented by a belief vector within a high dimensional probability simplex. It is well known that obtaining structural results for such high-dimensional POMDP is challenging. Firstly, we prove that for an infinite-horizon process subject to multiple competing assignable causes, a so-called conditional control limit policy is optimal. The optimal policy divides the belief space into two individually connected regions, which have analytical bounds. Next, we address a finite-horizon process with at least one absorbing state and show that a structured optimal policy can be established by transforming the belief space into a polar coordinate system, where a so-called polar control limit policy is optimal. Our model is general enough to include many existing models in the literature as special cases. The structural results also lead to significantly efficient algorithms for computing the optimal policies. In addition, we characterize the condition for some out-of-control state to be more desirable than the in-control state. The existence of such counterintuitive situation indicates that multi-state process control is drastically different from the two-state case.

Bayesian process control is a statistical process control (SPC) scheme that uses the posterior state probabilities as the control statistic. The key issue is to decide when to restore the process based on real-time observations. Such problems have been extensively studied in the framework of partially observable Markov decision processes (POMDP), with particular emphasis on the structure of optimal control policy. Almost all existing structural results on the optimal policies are limited to the two-state processes, where the class of control-limit policy is optimal. However, the two-state model is a gross simplification, as real production processes almost always involve multiple states. For example, a machine in the production system often has multiple failure modes differing in their effects; the deterioration process can often be divided into multiple stages with different degradation levels; the condition of a complex multi-unit system also requires a multi-state representation. We investigate the optimal control policies for multi-state processes with fixed sampling scheme, in which information about the process is represented by a belief vector within a high dimensional probability simplex. It is well known that obtaining structural results for such high-dimensional POMDP is challenging. Firstly, we prove that for an infinite-horizon process subject to multiple competing assignable causes, a so-called conditional control limit policy is optimal. The optimal policy divides the belief space into two individually connected regions, which have analytical bounds. Next, we address a finite-horizon process with at least one absorbing state and show that a structured optimal policy can be established by transforming the belief space into a polar coordinate system, where a so-called polar control limit policy is optimal. Our model is general enough to include many existing models in the literature as special cases. The structural results also lead to significantly efficient algorithms for computing the optimal policies. In addition, we characterize the condition for some out-of-control state to be more desirable than the in-control state. The existence of such counterintuitive situation indicates that multi-state process control is drastically different from the two-state case.

The application of Statistical Process Control (SPC) to software processes has been a challenging issue for software engineers and researchers. Although SPC is suggested for providing process control and achieving higher process maturity levels, there are very few resources that describe success stories, implementation details, and implemented guidelines for applying SPC to specific metrics. In this thesis the findings of a case study that is performed for investigating the applicability of SPC to software metrics in an emergent CMM Level 3 software organization are presented. As being one of the basic and most sophisticated tools of SPC, control charts are used for the analysis. The difficulties in application of Statistical Process Control to a CMM Level 3 organization are observed by using the existing data of defect density, rework percentage, productivity and review performance metrics and relevant suggestions are provided for dealing with them. Finally the analysis results are summarized and a guideline is prepared for software companies who want to utilize control charts by using their existing metric data.

Small and Medium Sized Enterprises (SMEs) were founded to improve the activity and effectiveness of small industries, to provide economic and social needs of the country, to increase the competitive level of the country, and to establish integration in the industry. In today&rsquo / s competition conditions, SMEs should continuously improve themselves / otherwise, they could lose their market shares. One of the major problems encountered in Turkish SMEs is poor quality activities / especially, not being able to exploit the Statistical Process Control (SPC) techniques. Production runs become shorter and shorter, and the product variety seems to be ever increasing, which cause short production runs. Using traditional control charts for short production runs can yield wrong and costly results. Instead of traditional control charts, short run charts such as Difference Charts (DNOM), Zed Charts, and Zed-Star Charts should be preferred.For this purpose, software that not only constructs short run control charts but also implements charts by tests to solve the problems of SMEs is developed. A Control Chart Selection Wizard, which is capable of emulating human expertise in finding a suitable control chart according to the user response for different cases is developed and added as a subprogram. Software was tested at Ar&ccedil / elik Dishwasher Plant in Ankara. The overall evaluation of the developed software, as regards the user, was satisfactory. The software can meet some requirements of the SMEs.

Routine pipette calibration is an essential part of any quality assurance and quality control program in the forensic sciences and beyond. Pipette calibration standards in a forensic laboratory are typically set to the limits outlined by the document ISO8655, published by the International Organization for Standardization for the general scientific community. Alternative methods exist that may be capable of monitoring pipette stability across time in a forensic setting. Statistical process control charts, or Shewhart charts, are one such form of process control, which is being investigated for its potential application to pipette calibration monitoring for forensic DNA laboratories. Indeed, the application of process control lines for monitoring the calibration of volumetric equipment is not without precedent.1 To investigate the applicability of process control charts for monitoring pipette stability, a series of X ̅ and S charts, a type of Shewhart chart, have been produced from eight years of collected calibration data. A total of 71 pipettes of the following sizes were examined: 1-10 µL, 1-10 µL multi-channel, 10-100 µL, 100-1000 µL, 1-3 µL, 30-300 µL, 5-50 µL, 5-50 µL multi-channel, and 500-5000 µL pipettes. The ISO8655 calibration recommended volume limits of these pipettes have been added to the charts for the purposes of comparison. With these charts, it is possible to assess pipette performance over time in comparison to the ISO8655 calibration standards and to the control limits imposed by the Shewhart charts. The completed charts suggest that the methodology proposed by Shewhart shows promise as a supplement to ISO8655 recommendations for monitoring pipette stability across time. To corroborate the value of using Shewhart charts to monitor pipette performance, a serial dilution study in conjunction with a series of simulations with dynamic modeling software was performed. This dilution study investigated whether the systematic biases shown by the Shewhart charts could be measured in a laboratory setting. The simulations investigated multiple hypothetical pipetting scenarios concerning various levels of systematic bias. The simulations consistently corroborated the value of Shewhart charts to enforce better compliance between a pipette’s nominal and actual volume delivery, while the serial dilution study offered partial evidence of systematic pipetting bias.

It is a difficult challenge to develop a feedback control system for Statistical Process Control (SPC) because there is no effective method that can be used to calculate the accurate magnitude of feedback control actions in traditional SPC. Suitable feedback adjustments are generated from the experiences of process engineers. This drawback means that the SPC technique can not be directly applied in an automatic system. This thesis is concerned with Fuzzy Sets and Fuzzy Logic applied to the uncertainty of relationships between the SPC (early stage) alarms and SPC implementation. Based on a number of experiments of the frequency distribution for shifts of abnormal process averages and human subjective decision, a Fuzzy-SPC control system is developed to generate the magnitude of feedback control actions using fuzzy inference. A simulation study which is written in C++ is designed to implement a Fuzzy-SPC controller with satisfactory results. To further reduce the control errors, a NeuroFuzzy network is employed to build NNFuzzy- SPC system in MATLAB. The advantage of the leaning capability of Neural Networks is used to optimise the parameters of the Fuzzy- X and Fuzzy-J? controllers in order to obtain the ideal consequent membership functions to adapt to the randomness of various processes. Simulation results show that the NN-Fuzzy-SPC control system has high control accuracy and stable repeatability. To further improve the practicability of a NN-Fuzzy-SPC system, a combined forecaster with EWMA chart and digital filter is designed to reduce the NN-Fuzzy-SPC control delay. For the EWMA chart, the smoothing constant 0 is investigated by a number of experiments and optimised in the forecast process. The Finite Impulse Response (FIR) lowpass filter is designed to smooth the input data (signal) fluctuations in order to reduce the forecast errors. An improved NN-Fuzzy-SPC control system which shows high control accuracy and short control delay can be applied in both automatic control and online quality control.

Com o aumento da produção mundial em proporções cada vez maiores, os processos industriais têm se tornando um desafio pela complexidade do seu gerenciamento. A identificação rápida e precisa de não conformidades é cada vez mais necessária e mais difícil de ser realizada. Este estudo propõe a integração do Controle de Engenharia com o Controle Estatístico de Processo, no monitoramento e controle de processos industriais, almejando a percepção mais rápida de anormalidades, visando à redução de problemas de especificação de produtos. Uma forma de autoajuste do controlador Proporcional-Integral e Derivativo (PID) é proposta, aumentando a robustez do sistema, empregando-se técnicas comumente utilizadas nos processos. O modelo matemático do processo, equacionando as relações das variáveis envolvidas, é estabelecido para determinação e especificação do controlador e de forma conjunta as cartas de controle são configuradas. O controlador projetado para a situação normal de operação atua no sentido de manter as variáveis de saída (controladas) dentro de especificações através do conhecimento de sua relação com as de entrada e de processo. As cartas de controle baseadas em modelos, monitorando os resíduos provenientes de ajustes de modelos ARIMA, acompanham as variações do processo, evitando variabilidade excessiva e possibilitando a detecção de comportamentos anormais, inclusive monitorando o desempenho do próprio controlador. Com a sinalização das cartas de controle, é realizada uma interferência na equação de ajuste do controlador. Empregando-se a simulação numérica, analisam-se os comportamentos do controlador e este, combinado com as cartas de controle. Falhas inseridas propositalmente, em cada variável controlada, foram devidamente sinalizadas. Estas sinalizações ocorreram mesmo em situações em que as variáveis estiveram mantidas dentro da especificação pelo controlador. Para o sistema de autoajuste, o aumento dos ganhos de contribuição das cartas de controle proporcionou maior acurácia das variáveis controladas (monitoradas). A integração proposta apresentou melhores resultados, quanto à manutenção das variáveis de saída próximas aos seus alvos, quando comparada com o controlador operando isoladamente. / With the increase of world production at bigger and bigger proportions, the industrial processes have become a challenge by the complexity of their management. The fast and precise identification of non-conformities is increasingly necessary and more difficult to be performed, preferably even before problems with product specification or waste can be considered. This study proposes the integration of Engineering Process Control with the Statistical Process Control, in monitoring and controlling of industrial processes, aiming the quicker perception of abnormalities, looking for the reduction of products specification problems. A form of self-adjustment of the Proportional-Integral and Derivative Controller (PID) is proposed, increasing the system robustness applying techniques commonly used in the processes. The mathematical model of the process, equating the relationship of the variables involved, is established to the determination and specification of the controller, and the control charts are configure in an integrated way. The controller projected to the normal operation situation, acts in the sense of keeping the exit variables (controlled) within the specifications through the knowledge of its relationship with that of the entrance and that of the process. The control charts based in models, monitoring the residues coming from the models adjustment ARIMA, follow up the process variations avoiding excessive variability and making it possible the detection of abnormal behaviors, even monitoring the performance of the controller itself. With the signalling of the control charts, an interference in the equation of adjustment of the controller is performed. Applying the numerical simulation, the controller behaviors are analyzed and this combined with the control charts. Intentionally inserted failures, in each controlled variable, were properly signalized. These signallings have happened even in situations where the variables were kept by the controller within the specification.. To the self-adjustment system, the increase of contribution gains of the control charts has provided greater accuracy of the controlled variables. The integration proposed has presented better results, in relation to maintain of the exit variables next to their targets, when compared to the controller operating in isolation.

A incerteza de medição ainda é tema de estudo e desenvolvimento, mesmo quase vinte anos após o lançamento do ISO GUM, que definiu um padrão internacional para o cálculo de incerteza. Mesmo assim, poucos estudos tratam da interface entre a incerteza de medição e outros aspectos da qualidade industrial, em especial ferramentas de controle de processo como cartas de controle. Assim, o objetivo deste trabalho é estudar os efeitos da incerteza em cartas de controle e identificar sob quais situações de aplicação ela exerce maior influência. Foi feito um mapeamento dos artigos publicados entre 2004 e 2010 sobre incerteza de medição, para identificar quais as metodologias de cálculo e abordagem são mais utilizadas. Na sequência, foi desenvolvido um método incluir nas cartas de controle a incerteza do sistema de medição, modificando as probabilidades de erros tipo I e tipo II. Finalmente, foi elaborada uma simulação computacional utilizando o método desenvolvido, para que fosse possível analisar quais as situações nas quais a incerteza de medição deve ser obrigatoriamente considerada e quais ela pode ser negligenciada. / The measurement uncertainty is still subject of study and development, even almost twenty years after the release of the ISO GUM, which set an international standard for the its calculation. Still, few studies have addressed the interface between the measurement uncertainty and other aspects of industrial quality, in particular process control tools like control charts. The objective of this work is to study the effects of measurement uncertainty on control charts and to identify the situations in which its application exerts greater influence. A research was made, mapping of the articles published between 2004 and 2010 regarding measurement uncertainty, aiming to identify which calculation methods and approaches are most widely used. After, a method was developed to include uncertainty of the measurement system in control charts, modifying the probabilities of type I and type II errors. Finally, it was created a computer simulation using the method developed, to make it possible to analyze those situations in which the measurement uncertainty must be considered and which it can be neglected.

Esta dissertação aborda a redução das perdas de qualidade em um processo de fabricação de fraldas descartáveis, através da implantação da ferramenta de Controle Estatístico de Processo (CEP), que tem por objetivo analisar e monitorar as principais variáveis e características de qualidade que influenciam no processo produtivo. Depois de implantada a ferramenta do CEP, são propostos alguns indicadores de desempenho para a área de produção da empresa que servirão de base para medir e acompanhar o desempenho de alguns processos, importantes para se atingir e acompanhar as metas de melhoria de qualidade almejadas pela empresa. A dissertação apresenta uma revisão detalhada da literatura abordando a qualidade, as perdas existentes no processo produtivo, o Desdobramento da Função Qualidade (QFD - Quality Function Deployment), o Controle Estatístico do Processo (CEP) e os indicadores de desempenho. Os resultados obtidos através da implantação do CEP mostram que a empresa deve investir na tecnologia do seu processo produtivo, pois em muitos aspectos sua qualidade esbarra na falta da mesma, mas deve se deter também na substituição de alguns fornecedores de suas matérias-primas, além de investir no treinamento de seus operadores, pois os mesmos devem estar constantemente renovando seus conhecimentos para prevenir falhas do processo. / This dissertation is about the reduction of the quality losses on a manufacturing process of diapers, through the implantation of the Statistical Process Control (SPC), which main purpose is to analyse and control the main variables and quality characteristics of the process that can inspire the improvement of the productive process. After introduce the tool, will be proposed some performance indicators for the production sector, just for accompanying and control some process in order that the company can reach it's purpose. The dissertation presents a specify revision o f the literature, approa~hing the quality, the losses in the productive process, the Quality Function Deployment (QFD), SPC and the performance indicators. The results gained with the implant of CEP show that the company should invest on the technology o f its manufacturing process, because in many ways the quality collides with a missing policy, but the company should also pay attention to the suppliers ofraw material and maybe replace them. Besides that, the company should invest on training its operator because they should be constantly renovating their knowledge to prevent failure on the process.