In the global automotive industry there is a drive toward integration of autonomous and human operated equipment. Monfared and Yang (2006:546) suggest that this dynamic requirement could be met with elements to be investigated in a research paper. Current investigations show a gap in management not having a guideline which can be used to help decide between automation versus human capital in the planning of new production facilities in the automotive assembly plant. (Skjerve and Skraaning, 2004:3). The purpose of this research is to determine what factors affect this decision-making process. In order to carry out this research, an in-depth literature review was conducted using various sources. The sources included, but were not limited to, interviews at assembly plants, the Nelson Mandela Metro University library, various e-journals and the internet. The literature review led to the finding of the factors which affect Levels of Automation (LOA) and to the development of the research instrument which was used to measure the impact of those factors. The results of fifty-two (52) respondents were then analysed and used as evidence to support the three hypotheses proposed. As a result of completing the above procedure the following hypotheses were supported. The greater the level of technology and the lower the skills of employees the greater the level of automation in an automotive assembly plant to be used. The greater the complexity of the assembly processes the lower the level of automation in an automotive assembly plant to be used. The higher the flexibility the greater the level of automation in an automotive assembly plant to be used. This means that managers and supervisors of assembly plants should consider the level of technology and skills of employees, flexibility and complexity during the design stages of an automotive assembly line as these factors will affect profitability by reducing waste, improve quality as well as allow for flexibility in customer demand in terms of volumes and product variance.

Abstract: Model-based control and optimization is the predominant paradigm in process systems engineering. The performance of model-based methods, however, rely heavily on the accuracy of the process model, which declines over the operation cycle due to various causes, such as catalyst deactivation, equipment aging, feedstock variability, and others. This work aims to tackle this challenge by considering two alternative approaches. The first approach replaces existing control and optimization methods with model-free reinforcement learning (RL). We apply a state-of-the-art reinforcement learning algorithm to a network of reactions, evaluate the performance of the RL controller in terms of setpoint tracking, disturbance rejection, and robustness to parameter uncertainties, and optimize the reward function to achieve the desired control and optimization performance. The second approach presents a novel framework for integrating Economic Model Predictive Control (EMPC) and RL for online model parameters estimation. In this framework, EMPC optimally operates the closed-loop system while maintaining closed-loop stability and recursive feasibility. At the same time, the RL agent continuously compares the measured state of the process with the model’s predictions, and modifies the model parameters accordingly to optimize the process. The performance of the proposed framework is illustrated on a network of reactions with challenging dynamics and practical significance.

No description available.

No description available.

The focus of this thesis is on building models by utilizing process information: from data, from our knowledge of physics, or both. The closer the model approximates reality, the better is the expected performance in forecasting, soft-sensing, process monitoring, optimization and advanced process control. In the domain of batch and continuous manufacturing, quality models can help in ensuring tightly controlled product quality, having safe and reliable operating conditions and reducing production/operation costs. To this end, first a parallel grey box model was built which makes use of a mechanistic model, and a subspace identification model for modeling a batch poly methyl methacrylate (PMMA) polymerisation process. The efficacy of such a parallel hybrid model in the context of a control problem was illustrated thereafter for reducing the volume of fines. Real-time implementation of models in many cases demand the model to be tractable and simple enough, and thus the parallel hybrid model was next adapted to have a linear representation, and then used for control computations. While the parallel hybrid modelling strategy shows great advantages in many applications, there can be other avenues of using fundamental process knowledge in conjunction with historical data. In one such approach, a unique way of adding mechanistic knowledge to improve the estimation ability of PLS models was proposed. The predictor matrix of PLS was augmented with additional trajectory information coming strategically from a mechanistic model. This augmented model was used as a soft-sensor to estimate batch end quality for a seeded batch crystallizer process. In a collaborative work with an industrial partner focusing on estimating important variables of a hydroprocessing unit, an operational data based input-output model was chosen as the right fit in the absence of available mechanistic knowledge. The usefulness of linear dynamic modeling tools for such applications was demonstrated. / Thesis / Doctor of Philosophy (PhD)

No description available.

This thesis focuses on the economic design of process control procedures for attributes sampling. The process is modeled as a continuous time, discrete space stochastic process which possesses the Markov property, and hence a Markov chain is used to describe its behavior. Two models are developed. The first model has fixed values of the decision variables and is optimized using the pattern search procedure. The second model is a dynamic formulation. The optimal decision policies developed using this model vary with the expected state of the process. Several cost components are considered in the mathematical development of each model. They are: the cost of sampling, the cost of process adjustment, and the cost of producing a defective unit. The cost of a false indication of the process state is also included in the fixed parameter model. Computer programs, written in Fortran IV are developed and used to find the optimal system designs. Example problems are presented to illustrate both of the models. The dynamic programming model is shown to offer considerable economic improvement over the steady state model in all of the examples. / Master of Science

A method for multivariate quality control with the dual objectives of providing a true level of sampling error probabilities for the joint control of several quality variables while also giving problem diagnoses for the quality variables individually. The method is comprised of an afine transformation of the multiple quality variables which creates a univariate test statistic used to monitor the quality and provide problem diagnoses. In practice, realized values of this statistic would be plotted as a time series on a control chart with multiple diagnosis intervals. For the analysis of the method’s effectiveness, the quality variables are assumed to be independent and normally distributed. The method is shown to be successful in achieving desired sampling error probabilities for any m quality variables in the case of positive shifts in the means of the variables. A second transformed variable is added for the diagnosis of shifts of unrestricted direction, and its effectiveness is analyzed. The sample size requirement of the afine transformation method is compared to the total sample size necessary when a separate Shewhart chart for the mean is maintained for each quality variable with the same overall sampling plan objectives. The power of the method to detect quality problems in general while disregarding specific diagnoses is compared to the power of Hotelling’s T² test for multivariate quality control. A comprehensive evaluation of the relative worth of the two methods is not possible since the T² statistic does not consider diagnoses of the individual quality variables. / Master of Science

Hybrid systems combine the continuous behavior evolution specified by differential equations with discontinuous changes specified by discrete event logic. Usually these systems in the processing industry can be identified as having to depend on discrete decisions regarding their operation. In process control there therefore is a challenge to automate these decisions. A model predictive control (MPC) strategy was proposed and verified for the control of hybrid systems. More specifically, the dynamic matrix control (DMC) framework commonly used in industry for the control of continuous variables was modified to deal with mixed integer variables, which are necessary for the modelling and control of hybrid systems. The algorithm was designed and commissioned in a closed control loop comprising a SCADA system and an optimiser (GAMS). GAMS (General Algebraic Modelling System) is an optimisation package that is able to solve for integer/continuous variables given a model of the system and an appropriate objective function. Online and offline closed loop tests were undertaken on a benchmark interacting tank system and a heating/cooling circuit. The algorithm was also applied to an industrial problem requiring the optimal sequencing of coal locks in real time. To complete the research concerning controller design for hybrid behavior, an investigation was undertaken regarding systems that have different modes of operation due to physicochemical (inherent) discontinuities e.g. a tank with discontinuous cross sectional area, fitted with an overflow. The findings from the online tests and offline simulations reveal that the proposed algorithm, with some system specific modification, was able to control each of the four hybrid systems under investigation. Based on which hybrid system was being controlled, by modifying the DMC algorithm to include integer variables, the mixed integer predictive controller (MIPC) was employed to initiate selections, switchings and determine sequences. Control of the interacting tank system was focused on an optimum selection in terms of operating positions for process inputs. The algorithm was shown to retain the usual features of DMC (i.e. tuning and dealing with multivariable interaction). For a system with multiple modes of operation i.e. the heating/cooling circuit, the algorithm was able to switch the mode of operation in order to meet operating objectives. The MPC strategy was used to good effect when getting the algorithm to sequence the operation of several coal locks. In this instance, the controller maintained system variables within certain operating constraints. Furthermore, soft constraints were proposed and used to promote operation close to operating constraints without the danger of computational failure due to constraint violations. For systems with inherent discontinuities, a MPC strategy was proposed that predicted trajectories which crossed discontinuities. Convolution models were found to be inappropriate in this instance and state space equations describing the dynamics of the system were used instead. / Thesis (M.Sc.Eng.)-University of Natal, Durban, 2002.

In this thesis, we first investigate the use of automated variance estimators in distribution-free statistical process control (SPC) charts for univariate autocorrelated processes. We introduce two variance estimators---the standardized time series overlapping area estimator and the so-called quick-and-dirty autoregressive estimator---that can be obtained from a training data set and used effectively with distribution-free SPC charts when those charts are applied to processes exhibiting nonnormal responses or correlation between successive responses. In particular, we incorporate the two estimators into DFTC-VE, a new distribution-free tabular CUSUM chart developed for autocorrelated processes; and we compare its performance with other state-of-the-art distribution-free SPC charts. Using either of the two variance estimators, the DFTC-VE outperforms its competitors in terms of both in-control and out-of-control average run lengths when all the competing procedures are tested on the same set of independently sampled realizations of selected autocorrelated processes with normal or nonnormal noise components. Next, we develop WDFTC, a wavelet-based distribution-free CUSUM chart for detecting shifts in the mean of a high-dimensional profile with noisy components that may exhibit nonnormality, variance heterogeneity, or correlation between profile components. A profile describes the relationship between a selected quality characteristic and an input (design) variable over the experimental region. Exploiting a discrete wavelet transform (DWT) of the mean in-control profile, WDFTC selects a reduced-dimension vector of the associated DWT components from which the mean in-control profile can be approximated with minimal weighted relative reconstruction error. Based on randomly sampled Phase I (in-control) profiles, the covariance matrix of the corresponding reduced-dimension DWT vectors is estimated using a matrix-regularization method; then the DWT vectors are aggregated (batched) so that the nonoverlapping batch means of the reduced-dimension DWT vectors have manageable covariances. To monitor shifts in the mean profile during Phase II operation, WDFTC computes a Hotelling's T-square--type statistic from successive nonoverlapping batch means and applies a CUSUM procedure to those statistics, where the associated control limits are evaluated analytically from the Phase I data. We compare WDFTC with other state-of-the-art profile-monitoring charts using both normal and nonnormal noise components having homogeneous or heterogenous variances as well as independent or correlated components; and we show that WDFTC performs well, especially for local shifts of small to medium size, in terms of both in-control and out-of-control average run lengths.