Repository logo
 

Faculty of Engineering and Built Environment

Permanent URI for this communityhttp://ir-dev.dut.ac.za/handle/10321/9

Browse

Subject: search.filters.subject.Process control

Search Results

Now showing 1 - 4 of 4
  • Thumbnail Image
    Item
    Assessment of control loop performance for nonlinear process
    (2017) Pillay, Nelendran; Govender, Poobalan
    Controller performance assessment (CPA) is concerned with the design of analytical tools that are utilized to evaluate the performance of process control loops. The objective of the CPA is to ensure that control systems operate at their full potential, and also to indicate when a controller design is performing unsatisfactorily under current closed loop conditions. Such monitoring efforts are imperative to minimize product variability, improve production rates and reduce wastage. Various studies conducted on process control loop performance indicate that as many as 60% of control loops often suffer from some kind of performance problem. It is therefore an important task to detect unsatisfactory control loop behavior and suggest remedial action. Such a monitoring system must be integrated into the control system life span as plant changes and hardware issues become apparent. CPA is well established for linear systems. However, not much research has been conducted on CPA for nonlinear systems. Traditional CPA analytical tools depend on the theoretical minimum variance control law that is derived from models of linear systems. In systems exhibiting dominant nonlinear behavior, the accuracy of linear based CPA is compromised. In light of this, there is a need to broaden existing CPA knowledge base with comprehensive benchmarking indices for the performance analysis of nonlinear process control systems. The research efforts presented in this thesis focuses on the development and analysis of such CPA tools for univariate nonlinear process control loops experiencing the negative effects of dominant nonlinearities emanating from the process. Two novel CPA frameworks are proposed; first a model based nonlinear assessment index is developed using an open loop model of the plant in an artificial neural network NARMAX (NNARMAX) representation. The nonlinear control loop is optimized offline using a proposed Nelder Mead-Particle Swarm Optimization (NM-PSO) hybrid search to determine global optimal control parameters for a gain scheduled PID controller. Application of the benchmark in real-time utilizes a synthetic process output derived from the NNARMAX system which is compared to the actual closed loop performance. In the case where no process model is available, a second method is presented. An autonomous data driven approach based on Multi-Class Support Vector Machines (MC- SVMs) is developed and analyzed. Unlike the model based method, the closed loop performance is classified according to five distinct class groups. MC-SVM classifier requires minimal process loop information other than routine operating closed loop data. Several simulation case studies conducted using MATLAB™ software package demonstrate the effectiveness of the proposed performance indices. Furthermore, the methodologies presented in this work were tested on real world systems using control loop data sets from a computer interfaced full scale pilot pH neutralization plant and pulp and paper industry.
  • Thumbnail Image
    Item
    Neural network modelling and prediction of the flotation deinking behaviour of industrial paper recycling processes
    (Nordic Pulp & Paper Research Journal, 2014) Pauck, Walter James; Venditti, Richard; Pocock, Jon; Andrew, Jerome Edward
    The removal of ink from recovered papers by flotation deinking is considered to be the “heart” of the paper recycling process. Attempts to model the deinking flotation process from first principles has resulted in complex and not readily usable models. Artificial neural networks are adept at modelling complex and poorly understood phenomena. Based on data generated in a laboratory, artificial neural network models were developed for the flotation deinking process. Representative samples of recycled newsprint, magazines and fine papers were pulped and deinked by flotation in the laboratory, under a wide variety of practical conditions. The brightness, residual ink concentration and the yield were measured and used to train artificial neural networks. Regressions of approximately 0.95, 0.85 and 0.79 respectively were obtained. These models were validated using actual plant data from three different deinking plants manufacturing seven different grades of recycled pulp. It was found that the brightness and residual ink concentration could be predicted with correlations in excess of 0.9. Lower correlations of ca. 0.43 were obtained for the flotation yield. It is intended to use the data to develop predictive models to facilitate the management and optimization of commercial flotation deinking processes with respect to recycled paper inputs and process conditions.
  • Thumbnail Image
    Item
    Neural networks approach to process control : the case of processes with long dead times
    (1999) McLeod, Charles Meredith; Bajic, Vladimir B.
    This study relates to applications of static artificial neural networks (ANNs) to two basic problems of process control: (a) process model identification, and (b) optimal controller tuning. The emphasis is on model identification, where several novel techniques are introduced. A review of the use of ANNs for determining optimal controller settings is included as a logical adjunct which would make the complete system suitable for realisation as a portable or networked system. Three methods for obtaining good approximations for the parameters of first-order processes with long dead time using artificial neural networks (ANNs) are proposed and described. These are termed in this study: time-domain, frequency-domain and model-based methods. In each case the aim was to develop a brief one-shot test that could be applied with minimal disturbance to a closed loop control system. These methods build on existing techniques, but introduce the following novel aspects: 2. The frequency-domain method makes use of the first 81 components of the FFT without further selection as input to a static ANN to yield process parameter estimates. 3. The model-based method uses a simple single-neuron implementation of an ARX model and uses a static ANN to relate process parameter values to the weights of this neuron. In making the analysis, the process input and output are applied repetitively to the neuron model with delays getting progressively larger. Useful effects arising from this are explored. A technique in which ANN training sets are slightly distorted in a random way during training of a radial basis function is developed as part of the time- and frequencydomain methods. The benefits arising from this technique are demonstrated. These experimental ANN-based control methods are evaluated by means of simulations in which accuracy in the presence of measurement noise and performance with higher order processes is measured and analysed. Although the main theme of this study is first-order-plus-dead-time (FOPDT) processes, the full autotuning scheme is tested with some representative higher order processes. Finally, the composition of a complete autotuning scheme is proposed which includes the automatic generation of controller parameters by means of ANN s.
  • Thumbnail Image
    Item
    Design and development of a process control valve diagnostic system based on artificial neural network ensembles
    (2016) Sewdass, Sugith; Govender, Poobalan
    This research discusses the design and development of a computational intelligent based diagnostic system to assess the operating state of a process control valve. Process control valves react to a controller signal and are the main source of faults in a control loop. The elasticity inherent within a valve’s mechanical construction makes it prone to nonlinearities such as backlash, hysteresis and stiction. These nonlinearities negatively affect the performance of a process control loop during a control session. The diagnostic system proposed in this research utilises artificial neural network systems configured as ensembles to classify common control valve faults. Each ensemble functions as a ‘specialist’ trained to identify a specific loop fault. The team of specialized artificial neural networks are configured into a single comprehensive system to detect common control loops problems such as valve hysteresis, backlash, stiction and low air supply. The detection of a specific type of fault is achieved by comparing the mean square error output from each network. The ensemble having the lowest mean square error is the network that has been trained to identify a specific type of fault. Two practical methods to simulate control valve stiction and hysteresis are also presented in this study. These methods make it possible for researchers to investigate dynamics of nonlinear behaviour when these nonlinear effects occur in the control channel.