This thesis is mainly concerned with the estimation of parameters in autoregressive models with censored data. For convenience, attention is restricted to the first-order stationary autoregressive (AR(1)) model in which the response random variables are subject to right-censoring. In their present form, currently available methods of estimation in regression analysis with censored autocorrelated data, which includes the MLE, are applicable only if the errors of the AR component of the model are Gaussian. Use of these methods in AR processes with non-Gaussian errors requires, essentially, rederivations of the estimators. Hence, in this thesis, we propose new estimators which arerobust in the sense that they can be applied with minor or no modifications to AR models with non-Gaussian. We propose three estimators, two of which the form of the distribution of the errors needs to be specified. The third estimator is a distribution-free estimator. As the reference to this estimator suggests, it is free from distributional assumptions in the sense that the error distribution is calculated from the observed data. Hence, it can be used in a wide variety of applications.In the first part of the thesis, we present a summary of the various currently available estimators for the linear regression model with censored independent and identically distributed (i.i.d.) data. In our review of these estimators, we note that the linear regression model with censored i.i.d. data has been studied quite extensively. Yet, use of autoregressive models with censored data has received very little attention. Hence, the remainder of the thesis focuses on the estimation of parameters for censored autocorrelated data. First, as part of the study, we review currently available estimators in regression with censored autocorrelated data. Then we present descriptions of the new estimators for censored ++ / autocorrelated data. With the view that extensions to the AR(p), model, p > 1, and to left-censored data can be easily achieved, all the estimators, both currently available and new, are discussed in the context of the AR(1) model. Next, we establish some asymptotic results for the estimators in which specification of the form of the error distribution is necessary. This is followed by a simulation study based on Monte Carlo experiments in which we evaluate and compare the performances of the new and currently available estimators among themselves and with the least-squares estimator for the uncensored case. The performances of the asymptotic variance estimators of the parameter estimators are also evaluated.In summary, we establish that for each of the two new estimators for which the distribution of the errors is assumed known, under suitable conditions on the moments of the error distribution function, if the estimator is consistent, then it is also asymptotically normally distributed. For one of these estimators, if the errors are Gaussian and alternate observations are censored, then the estimator is consistent. Hence, for this special case, the estimator is consistent and asymptotically normal. The simulation results suggest that this estimator is comparable with the distribution-free estimator and a currently available pseudolikelihood (PL) estimator. All three estimators perform worse than the least squares estimator for the uncensored case. The MLE and another currently available PL estimator perform comparably not only with the least squares estimator for the uncensored case but also with estimators from the abovementioned group of three estimators, which includes the distribution-free estimator. The other new estimator for which the form of the error distribution is assumed known compares favourably with the least- squares estimator for the uncensored case ++ / and better than the rest of the estimators when the true value of the autoregression parameter is 0.2. When the true value of the parameter is 0.5, this estimator performs comparably with the rest of the estimators and worse when the true value of the parameter is O.S. The simulation results of the asymptotic variance estimators suggest that for each estimator and for a fixed value of the true autoregression parameter, if the error distribution is fixed and the censoring rate is constant, the asymptotic formulas lead to values which are asymptotically insensitive to the censoring pattern. Also, the estimated asymptotic variances decrease as the sample size increases and their behaviour, with respect to changes in the true value of autoregression parameter, is consistent with the behaviour of the asymptotic variance of the least-squares estimator for the uncensored case.Some suggestions for possible extensions conclude the thesis.

Most products are produced by several process steps and have more than one interested quality characteristics. If each step of the process is independent, and the observations taken from the process are also independent then we may use Shewhart control chart at each step. However, in many processes, most production steps are dependent and the observations taken from the process are correlated. In this research, we consider the process has two dependent steps and the observations taken from the process are correlated over time. We construct the individual residual control chart to monitor the previous process and the cause-selecting control chart to monitor the current process. Then simulate all the states occur in the process and present the individual residual control chart and the cause-selecting control chart of the simulations. Furthermore compare the proposed control charts with the Hotelling T2 control chart. At last, we give an example to illustrate how to construct the proposed control From the proposed control charts, we can determine which step of the process is out of control easily. If there is a signal in the individual residual control chart, it means the previous process is out of control. If there is a signal in the cause-selecting control chart, it means the current process is out of control. The Hotelling T2 control chart only indicate the process is out of control but does not detect which step of the process is out of control.

<p>The use of process capability indices in the industry is traditionally based on the assumptions that a sample from a process are observations on independently, identically and normally distributed random variables<i>, IIN</i>. However, all three assumptions are open to discussion and in this thesis, the estimation of the indices is studied when the assumption of independence is not fulfilled.</p><p>In five reports, the indices <i>C</i><i>pk</i> and <i>C</i><i>pm </i>are studied, and instead of random samples, samples are regarded as observations on a time series.</p><p>In the first four reports, each index is studied for either an <i>AR(1)</i> or an <i>MA(1)</i> process and the fifth report, both indices are studied for a general <i>ARMA(p,q</i>) process.</p><p>In all reports, alternatives to <i>C</i><i>pk</i><i> </i>and <i>C</i><i>pm</i><i> </i>are suggested as well as point and interval estimators for the suggested indices. The accuracy of interval estimators are evaluated through large Monte Carlo simulations and the difference between empirical coverage rates and nominal confidence limits are calculated.</p><p>It was found in all reports that a dependency among observations has a great impact on the coverage rates. The coverage rate difference depends on both the size of the autocorrelation and the type of time series model and for the original <i>C</i><i>pk</i> and <i>C</i><i>pm</i> the difference can be large. With the suggested alternative indices, however, the differences are always reduced and unless the autocorrelations are close to ±1, the sizes of differences are of little consequence.</p>

The use of process capability indices in the industry is traditionally based on the assumptions that a sample from a process are observations on independently, identically and normally distributed random variables, IIN. However, all three assumptions are open to discussion and in this thesis, the estimation of the indices is studied when the assumption of independence is not fulfilled. In five reports, the indices Cpk and Cpm are studied, and instead of random samples, samples are regarded as observations on a time series. In the first four reports, each index is studied for either an AR(1) or an MA(1) process and the fifth report, both indices are studied for a general ARMA(p,q) process. In all reports, alternatives to Cpk and Cpm are suggested as well as point and interval estimators for the suggested indices. The accuracy of interval estimators are evaluated through large Monte Carlo simulations and the difference between empirical coverage rates and nominal confidence limits are calculated. It was found in all reports that a dependency among observations has a great impact on the coverage rates. The coverage rate difference depends on both the size of the autocorrelation and the type of time series model and for the original Cpk and Cpm the difference can be large. With the suggested alternative indices, however, the differences are always reduced and unless the autocorrelations are close to ±1, the sizes of differences are of little consequence.

Statistical Process Control (SPC), and in particular control charting, is widely used to achieve and maintain control of various processes in manufacturing. A control chart is a graphical display that plots quality characteristics versus the sample number or the time line. Interest in effective implementation of control charts for autocorrelated processes has increased in recent years. However, because of the complexities involved, few systematic design approaches have thus far been developed. Many control charting methods can be viewed as the charting of the output of a linear filter applied to the process data. In this dissertation, we generalize the concept of linear filters for control charts and propose new control charting schemes, the general linear filter (GLF) and the 2nd-order linear filter, based on the generalization. In addition, their optimal design methodologies are developed, where the filter parameters are optimally selected to minimize the out-of-control Average Run Length (ARL) while constraining the in-control ARL to some desired value. The optimal linear filters are compared with other methods in terms of ARL performance, and a number of their interesting characteristics are discussed for various types of mean shifts (step, spike, sinusoidal) and various ARMA process models (i.i.d., AR(1), ARMA(1,1)). Also, in this work, a new discretization approach for substantially reducing the computational time and memory use for the Markov chain method of calculating the ARL is proposed. Finally, a gradient-based optimization strategy for searching optimal linear filters is illustrated.

有鑑於資產報酬常具有自我相關的特性，本文探討當標的資產報酬服從一階移動平均過程之選擇權（MA(1)-type option）評價。研究結果顯示，除了總變異因子（total volatility input）不同外，MA(1)-type option 的評價公式與 Black and Scholes 模型極為相似。而根據數值分析的結果，即使資產報酬間自我相關的程度薄弱，由一階移動平均過程產生之自我相關仍會對選擇權價值造成顯著影韾。 / This paper derives the closed-form formula for a European option on an asset with returns following a continuous-time type of first-order moving average process, which is named as an MA(1)-type option. The pricing formula of these options is similar to that of Black and Scholes except for the total volitility input. Specifically, the total volatility input of MA(1)-type options is the conditional standard deviation of continuous-compounded returns over the option's remaining life, whereas the total volatility input of Black and Scholes is indeed the diffusion coefficient of a geometric Brownian motion times the square root of an option's time to maturity. Based on the result of numerical analyses, the impact of autocorrelation induced by the MA(1)-type process is significant to option values even when the autocorrelation between asset returns is weak.

Im Kern der vorliegenden Arbeit wird eine neue Vorfahrtstrategie zur Steuerung von Materialflüssen an Zusammenführungen vorgestellt. Das Hauptanwendungsgebiet stellen innerbetriebliche Transportsysteme dar, wobei die Erkenntnisse auf beliebige Transport- bzw. Bediensysteme übertragbar sind. Die Arbeit grenzt sich mit der Annahme autokorrelierter Ankunftsprozesse von bisheriger Forschung und Entwicklung ab. Bis dato werden stets unkorrelierte Ströme angenommen bzw. findet keine spezielle Beachtung autokorrelierter Ströme bei der Vorfahrtsteuerung statt. Untersuchungen zeigen aber, dass zum einen mit hoher Konfidenz mit autokorrelierten Materialflüssen zu rechnen ist und in diesem Fall zum anderen von einem erheblichen Einfluss auf die Systemleistung ausgegangen werden muss. Zusammengefasst konnten im Rahmen der vorliegenden Arbeit 68 Realdatensätze verschiedener Unternehmen untersucht werden, mit dem Ergebnis, dass ca. 95% der Materialflüsse Autokorrelation aufweisen. Ferner wird hergeleitet, dass Autokorrelation intrinsisch in Materialflusssystemen entsteht. Die Folgen autokorrelierter Prozesse bestehen dabei in längeren Durchlaufzeiten, einem volatileren Systemverhalten und höheren Wahrscheinlichkeiten von Systemblockaden. Um die genannten Effekte an Zusammenführungen zu eliminieren, stellt die Arbeit eine neue Vorfahrtstrategie HAFI – Highest Autocorrelated First vor. Diese priorisiert die Ankunftsprozesse anhand deren Autokorrelation. Konkret wird die Vorfahrt zunächst so lange nach dem Prinzip First Come First Served gewährt, bis richtungsweise eine spezifische Warteschlangenlänge überschritten wird. Der jeweilige Wert ergibt sich aus der Höhe der Autokorrelation der Ankunftsprozesse. Vorfahrt bekommt der Strom mit der höchsten Überschreitung seines Grenzwertes. Die Arbeit stellt ferner eine Heuristik DyDeT zur automatischen Bestimmung und dynamischen Anpassung der Grenzwerte vor. Mit einer Simulationsstudie wird gezeigt, dass HAFI mit Anwendung von DyDeT die Vorzüge der etablierten Vorfahrtstrategien First Come First Served und Longest Queue First vereint. Dabei wird auch deutlich, dass die zwei letztgenannten Strategien den besonderen Herausforderungen autokorrelierter Ankunftsprozesse nicht gerecht werden. Bei einer Anwendung von HAFI zur Vorfahrtsteuerung können Durchlaufzeiten und Warteschlangenlängen auf dem Niveau von First Come First Served erreicht werden, wobei dieses ca. 10% unter dem von Longest Queue First liegt. Gleichzeitig ermöglicht HAFI, im Gegensatz zu First Come First Served, eine ähnlich gute Lastbalancierung wie Longest Queue First. Die Ergebnisse stellen sich robust gegenüber Änderungen der Auslastung sowie der Höhe der Autokorrelation dar. Gleichzeitig sind die Erkenntnisse unabhängig der Analyse einer isolierten Zusammenführung und der Anordnung mehrerer Zusammenführungen in einem Netzwerk. / The work at hand presents a novel strategy to control arrival processes at merges. The main fields of application are intralogistics transport systems. Nevertheless, the findings can be adapted to any queuing system. In contrast to further research and development the thesis assumes autocorrelated arrival processes. Up until now, arrivals are usually assumed to be uncorrelated and there are no special treatments for autocorrelated arrivals in the context of merge controlling. However, surveys show with high reliability the existence of autocorrelated arrivals, resulting in some major impacts on the systems\' performance. In detail, 68 real-world datasets of different companies have been tested in the scope of this work, and in 95% of the cases arrival processes significantly show autocorrelations. Furthermore, the research shows that autocorrelation comes from the system itself. As a direct consequence it was observed that there were longer cycle times, more volatile system behavior, and a higher likelihood of deadlocks. In order to eliminate these effects at merges, this thesis introduces a new priority rule called HAFI-Highest Autocorrelated First. It assesses the arrivals\' priority in accordance to their autocorrelation. More concretely, priority initially is given in accordance to the First Come First Served scheme as long as specific direction-wise queue lengths are not exceeded. The particular thresholds are determined by the arrival processes\' autocorrelation, wherein the process with the highest volume gets priority. Furthermore, the thesis introduces a heuristic to automatically and dynamically determine the specific thresholds of HAFI-so called DyDeT. With a simulation study it can be shown that HAFI in connection with DyDeT, combines the advantages of the well-established priority rules First Come First Served and Longest Queue First. It also becomes obvious that the latter ones are not able to deal with the challenges of autocorrelated arrival processes. By applying HAFI cycling times and mean queue lengths on the level of First Come First Served can be achieved. These are about 10% lower than for Longest Queue First. Concomitantly and in contrast to First Come First Served, HAFI also shows well balanced queues like Longest Queue First. The results are robust against different levels of throughput and autocorrelation, respectively. Furthermore, the findings are independent from analyzing a single instance of a merge or several merges in a network.

Im Kern der vorliegenden Arbeit wird eine neue Vorfahrtstrategie zur Steuerung von Materialflüssen an Zusammenführungen vorgestellt. Das Hauptanwendungsgebiet stellen innerbetriebliche Transportsysteme dar, wobei die Erkenntnisse auf beliebige Transport- bzw. Bediensysteme übertragbar sind. Die Arbeit grenzt sich mit der Annahme autokorrelierter Ankunftsprozesse von bisheriger Forschung und Entwicklung ab. Bis dato werden stets unkorrelierte Ströme angenommen bzw. findet keine spezielle Beachtung autokorrelierter Ströme bei der Vorfahrtsteuerung statt. Untersuchungen zeigen aber, dass zum einen mit hoher Konfidenz mit autokorrelierten Materialflüssen zu rechnen ist und in diesem Fall zum anderen von einem erheblichen Einfluss auf die Systemleistung ausgegangen werden muss. Zusammengefasst konnten im Rahmen der vorliegenden Arbeit 68 Realdatensätze verschiedener Unternehmen untersucht werden, mit dem Ergebnis, dass ca. 95% der Materialflüsse Autokorrelation aufweisen. Ferner wird hergeleitet, dass Autokorrelation intrinsisch in Materialflusssystemen entsteht. Die Folgen autokorrelierter Prozesse bestehen dabei in längeren Durchlaufzeiten, einem volatileren Systemverhalten und höheren Wahrscheinlichkeiten von Systemblockaden. Um die genannten Effekte an Zusammenführungen zu eliminieren, stellt die Arbeit eine neue Vorfahrtstrategie HAFI – Highest Autocorrelated First vor. Diese priorisiert die Ankunftsprozesse anhand deren Autokorrelation. Konkret wird die Vorfahrt zunächst so lange nach dem Prinzip First Come First Served gewährt, bis richtungsweise eine spezifische Warteschlangenlänge überschritten wird. Der jeweilige Wert ergibt sich aus der Höhe der Autokorrelation der Ankunftsprozesse. Vorfahrt bekommt der Strom mit der höchsten Überschreitung seines Grenzwertes. Die Arbeit stellt ferner eine Heuristik DyDeT zur automatischen Bestimmung und dynamischen Anpassung der Grenzwerte vor. Mit einer Simulationsstudie wird gezeigt, dass HAFI mit Anwendung von DyDeT die Vorzüge der etablierten Vorfahrtstrategien First Come First Served und Longest Queue First vereint. Dabei wird auch deutlich, dass die zwei letztgenannten Strategien den besonderen Herausforderungen autokorrelierter Ankunftsprozesse nicht gerecht werden. Bei einer Anwendung von HAFI zur Vorfahrtsteuerung können Durchlaufzeiten und Warteschlangenlängen auf dem Niveau von First Come First Served erreicht werden, wobei dieses ca. 10% unter dem von Longest Queue First liegt. Gleichzeitig ermöglicht HAFI, im Gegensatz zu First Come First Served, eine ähnlich gute Lastbalancierung wie Longest Queue First. Die Ergebnisse stellen sich robust gegenüber Änderungen der Auslastung sowie der Höhe der Autokorrelation dar. Gleichzeitig sind die Erkenntnisse unabhängig der Analyse einer isolierten Zusammenführung und der Anordnung mehrerer Zusammenführungen in einem Netzwerk.:1 Einleitung 1 1.1 Motivation 1 1.2 Zielsetzung, wissenschaftlicher Beitrag 4 1.3 Konzeption 5 2 Grundlagen 7 2.1 Automatisierung, Steuern, Regeln 7 2.2 System, Modell 10 2.3 Stochastik, Statistik 14 2.3.1 Wahrscheinlichkeitsverteilungen 14 2.3.2 Zufallszahlengeneratoren 21 2.3.3 Autokorrelation als Ähnlichkeits- bzw. Abhängigkeitsmaß 24 2.4 Simulation 29 2.5 Warteschlangentheorie und -modelle 32 2.6 Materialflusssystem 35 2.7 Materialflusssteuerung 37 2.7.1 Steuerungssysteme 37 2.7.2 Steuerungsstrategien 40 2.8 Materialflusssystem charakterisierende Kennzahlen 46 3 Stand der Forschung und Technik 51 3.1 Erzeugung autokorrelierter Zufallszahlen 51 3.1.1 Autoregressive Prozesse nach der Box-Jenkins-Methode 52 3.1.2 Distorsions-Methoden 54 3.1.3 Copulae 56 3.1.4 Markovian Arrival Processes 58 3.1.5 Autoregressive Prozesse mit beliebiger Randverteilung 61 3.1.6 Weitere Verfahren 64 3.1.7 Bewertung der Verfahren und Werkzeuge zur Generierung 65 3.2 Wirken von Autokorrelation in Bediensystemen 68 3.3 Fallstudien über Autokorrelation in logistischen Systemen 75 3.4 Ursachen von Autokorrelation in logistischen Systemen 89 3.5 Steuerung von Ankunftsprozessen an Zusammenführungen 96 3.6 Steuerung autokorrelierter Ankunftsprozesse 100 4 Steuerung autokorrelierter Ankunftsprozesse an Zusammenführungen 105 4.1 Modellannahmen, Methodenauswahl, Vorbetrachtungen 106 4.2 First Come First Served und Longest Queue First 114 4.3 Highest Autocorrelated First 117 4.3.1 Grundprinzip 117 4.3.2 Bestimmung der Grenzwerte 127 4.3.3 Dynamische Bestimmung der Grenzwerte mittels „DyDeT“ 133 4.4 Highest Autocorrelated First in Netzwerken 150 4.5 Abschließende Bewertung und Diskussion 161 5 Zusammenfassung und Ausblick 167 Primärliteratur 172 Normen und Standards 194 Abbildungsverzeichnis 197 Tabellenverzeichnis 199 Pseudocodeverzeichnis 201 Abkürzungsverzeichnis 203 Symbolverzeichnis 205 Erklärung an Eides statt 209 / The work at hand presents a novel strategy to control arrival processes at merges. The main fields of application are intralogistics transport systems. Nevertheless, the findings can be adapted to any queuing system. In contrast to further research and development the thesis assumes autocorrelated arrival processes. Up until now, arrivals are usually assumed to be uncorrelated and there are no special treatments for autocorrelated arrivals in the context of merge controlling. However, surveys show with high reliability the existence of autocorrelated arrivals, resulting in some major impacts on the systems\' performance. In detail, 68 real-world datasets of different companies have been tested in the scope of this work, and in 95% of the cases arrival processes significantly show autocorrelations. Furthermore, the research shows that autocorrelation comes from the system itself. As a direct consequence it was observed that there were longer cycle times, more volatile system behavior, and a higher likelihood of deadlocks. In order to eliminate these effects at merges, this thesis introduces a new priority rule called HAFI-Highest Autocorrelated First. It assesses the arrivals\' priority in accordance to their autocorrelation. More concretely, priority initially is given in accordance to the First Come First Served scheme as long as specific direction-wise queue lengths are not exceeded. The particular thresholds are determined by the arrival processes\' autocorrelation, wherein the process with the highest volume gets priority. Furthermore, the thesis introduces a heuristic to automatically and dynamically determine the specific thresholds of HAFI-so called DyDeT. With a simulation study it can be shown that HAFI in connection with DyDeT, combines the advantages of the well-established priority rules First Come First Served and Longest Queue First. It also becomes obvious that the latter ones are not able to deal with the challenges of autocorrelated arrival processes. By applying HAFI cycling times and mean queue lengths on the level of First Come First Served can be achieved. These are about 10% lower than for Longest Queue First. Concomitantly and in contrast to First Come First Served, HAFI also shows well balanced queues like Longest Queue First. The results are robust against different levels of throughput and autocorrelation, respectively. Furthermore, the findings are independent from analyzing a single instance of a merge or several merges in a network.:1 Einleitung 1 1.1 Motivation 1 1.2 Zielsetzung, wissenschaftlicher Beitrag 4 1.3 Konzeption 5 2 Grundlagen 7 2.1 Automatisierung, Steuern, Regeln 7 2.2 System, Modell 10 2.3 Stochastik, Statistik 14 2.3.1 Wahrscheinlichkeitsverteilungen 14 2.3.2 Zufallszahlengeneratoren 21 2.3.3 Autokorrelation als Ähnlichkeits- bzw. Abhängigkeitsmaß 24 2.4 Simulation 29 2.5 Warteschlangentheorie und -modelle 32 2.6 Materialflusssystem 35 2.7 Materialflusssteuerung 37 2.7.1 Steuerungssysteme 37 2.7.2 Steuerungsstrategien 40 2.8 Materialflusssystem charakterisierende Kennzahlen 46 3 Stand der Forschung und Technik 51 3.1 Erzeugung autokorrelierter Zufallszahlen 51 3.1.1 Autoregressive Prozesse nach der Box-Jenkins-Methode 52 3.1.2 Distorsions-Methoden 54 3.1.3 Copulae 56 3.1.4 Markovian Arrival Processes 58 3.1.5 Autoregressive Prozesse mit beliebiger Randverteilung 61 3.1.6 Weitere Verfahren 64 3.1.7 Bewertung der Verfahren und Werkzeuge zur Generierung 65 3.2 Wirken von Autokorrelation in Bediensystemen 68 3.3 Fallstudien über Autokorrelation in logistischen Systemen 75 3.4 Ursachen von Autokorrelation in logistischen Systemen 89 3.5 Steuerung von Ankunftsprozessen an Zusammenführungen 96 3.6 Steuerung autokorrelierter Ankunftsprozesse 100 4 Steuerung autokorrelierter Ankunftsprozesse an Zusammenführungen 105 4.1 Modellannahmen, Methodenauswahl, Vorbetrachtungen 106 4.2 First Come First Served und Longest Queue First 114 4.3 Highest Autocorrelated First 117 4.3.1 Grundprinzip 117 4.3.2 Bestimmung der Grenzwerte 127 4.3.3 Dynamische Bestimmung der Grenzwerte mittels „DyDeT“ 133 4.4 Highest Autocorrelated First in Netzwerken 150 4.5 Abschließende Bewertung und Diskussion 161 5 Zusammenfassung und Ausblick 167 Primärliteratur 172 Normen und Standards 194 Abbildungsverzeichnis 197 Tabellenverzeichnis 199 Pseudocodeverzeichnis 201 Abkürzungsverzeichnis 203 Symbolverzeichnis 205 Erklärung an Eides statt 209

[pt] Esta dissertação trata do controle estatístico de processos (CEP) multi-canal, assunto pouco tratado na literatura especializada. As técnicas encontradas na literatura pressupõem condições de validade nem sempre verificadas na prática, a saber: medidas sucessivas efetuadas em cada canal independentes e identicamente distribuídas; todos os canais ajustados, com mesma média e desvio-padrão; e (na maioria dos trabalhos) canais independentes, sem correlação cruzada. O estudo foi motivado por um caso real, em que nenhuma dessas condições se verifica. Este trabalho propõe adaptações e extensões de técnicas existentes para lidar com processos nessa situação; detalha como aplicá-las; ilustra sua aplicação no caso prático analisado; discute limitações e propõe alternativas, e inicia uma discussão sobre as diferentes condições (características das situações práticas) em que cada uma das alternativas é mais apropriada. O trabalho iniciou- se com uma análise exploratória dos processos da empresa, de modo a permitir um diagnóstico do CEP que vinha sendo realizado, e fundamentar a proposta de um novo esquema de controle, mais adequado. O esquema proposto, aplicado aos dados, sinalizou problemas com os processos que as técnicas empregadas não sinalizavam. Embora, em virtude dos prazos para finalização da dissertação e da programação da produção da empresa, que teve interrupções, não tenha sido possível incluir nesta análise a investigação de causas especiais, com revisão dos limites de controle e utilização dos gráficos assim revistos no monitoramento on line, para fins de acompanhamento do desempenho do esquema proposto, mesmo assim a aplicação desse esquema aos dados disponíveis demonstrou ser ele mais sensível a causas especiais que os gráficos que vinham sendo utilizados, e levantou algumas questões, não abordadas na literatura, que são indicadas para pesquisa futura. / [en] This dissertation tackles the problem of statistical control of multi-channel processes, which has been scarcely dealt with in the specialized literature. The techniques found in the literature assume validity conditions which are not always verified in practice, namely: successive measurements taken in each channel should be independent an identically distributed; all the channels should be adjusted, with same mean and standard deviation; and most works assume the channels to be independent, with no cross correlation. The study was motivated by a real case, in which no one of these conditions holds. This work proposes adaptations and extensions of existing techniques in order to deal with this and similar real situations; details the application of the adapted/extended techniques; illustrates their application in the case under analysis; discusses limitations and proposes alternatives, and initiates a discussion about the different conditions (practical situations´ characteristics) in which each alternative is most appropriate. The work began with an exploratory data analysis of the enterprise´s production processes, so as to enable a diagnosis of the statistical process control procedures that were in use, and serve as a basis for the proposal of a new, more adequate, control scheme. The proposed scheme, when applied to the data, signalled problems with the processes which the techniques in use did not signal. Due to the deadlines for ending this dissertation and to programmed interruptions in the production, it has not been possible to include in the analysis the search for special causes with corresponding revisions of the charts´ control limits and the use of the revised charts in on-line monitoring, for the purposes of feedback on the proposed scheme´s performance. Its application to the available data has nevertheless shown it more sensitive to special causes than the charts that were in use, and raised some issues not approached in the literature, which are left as indications for future research.