Adam Adgar

Performance Improvements at Surface Water Treatment Works Using ANN-Based Automation Schemes

Due to their ability to capture non-linear information very efficiently, artificial neural networks [ANNs] have found great popularity amongst the ‘control community’ and other disciplines. This paper discusses some recent applications of the ANNs at surface water treatment works. The range of application is quite diverse and covers modelling, simulation, condition monitoring, fault detection and control strategy design and implementation. Attempts to improve the performance of water treatment works through the application of improved control and measurement have had variable success. The most quoted reason for this is that the individual dynamic operations defining the treatment cycle are complex, highly non-linear and poorly understood. These problems are compounded by the use of faulty or badly maintained sensors. The efficient and robust operation of any industrial system is critically dependent on the quality of the measurements made. Also, the structure of the control policy and choice of the individual controller parameters are important decisions to the economic operation. Three examples are used to describe how the introduction of ANNs has resulted in more reliable system measurement and more efficient pH and coagulation control. A final example, shows an approach to the use of an ANN to provide ‘assistance’ to a conventional proportional-integral controller in the form of automatic on-line tuning of the controller parameters.

Cost effective water clarification control using a neurally informed control strategy

The abstraction, treatment and supply of potable (drinking) water presents a range of special problems. Early attempts at introducing and extending the range of automatic control at a water treatment plant were invariably compromised by the poor quality of the instrumentation available. Improved sensor technology and the advent of microprocessor based programmable controllers has allowed the implementation of nonstandard solutions. This paper describes the steps followed in developing an accurate feedforward automatic clarification control strategy. The existing strategy was occasionally compromised when a colour measurement was corrupted, usually under conditions of high turbidity. It should be emphasised that the colour monitor used is being asked to operate at and beyond the range of its original design specification. This occurs because there are no other alternative instruments. The paper shows how the use of an artificial neural network produces a solution to a seemingly intractable problem.

Integrating Statistical and Engineering Control Techniques

Statistical Process Control (SPC) is a set of statistical procedures that can be used to improve product quality and process productivity at reduced cost. The objective is to bring and keep processes in a state where any remaining variations are those inherent to the process. Traditionally, this is achieved by successive plotting and comparison of a chosen sample statistic with the appropriate control limits that are determined assuming independent and normally distributed random samples l . If the plotted variable exceeds the respective control limits, the process is considered to be out of statistical control. The procedure adopted in SPC charting methods is to try to separate variation that is ordinarily expected of a process, such as simple random variation, from that which may be due to the quality of raw materials or changes in operating conditions, etc. Corrective action is applied in the form of identification, elimination or compensation for these assignable causes of variation. There are guidelines on whether SPC or automatic control should be applied. These are generally based on the type of process and whether manipulation incurs costz.4. On the issue of costs, it has been suggested that SPC should be applied when costs are incurred in making manipulations. Conversely, automatic control should be employed when there is no cost associated with making process adjustments. Although automatic control technology can be found in the parts manufacturing industry, the assessment of final product quality is usually achieved by discrete inspection. SPC therefore fits nicely in such environments, where sampling procedures, measures of quality, diagnosis of process misbehaviour, quality control, etc. are embedded within a statis-

Challenges in the Development of an E-Maintenance System

Abstract e-Maintenance is generally understood to be the technology that makes the required information available for the maintenance engineer enabling communication with the supporting system irrespective of where the machine under inspection or maintenance actions is located. Increasing interest in improving overall production efficiency of manufacturing operations, together with technical developments in sensor and analysis equipment and wireless technology has led to accelerated research into e-Maintenance. Findings from a large European research project - Dynamite (Dynamic Decisions in Maintenance) are used to discuss the roles of various actors and roles in the proposed e-Maintenance scheme. The paper also highlights the main challenges faced in developing fully functional and economically optimised e-Maintenance support systems for modern production environments.

Simulation Software Accelerates Development of Neural Network Control Applications in the Water Industry

Artificial neural network (ANN) applications have become popular in the field of water treatment process for modelling and control purposes. This is due to the fact that the processes are not well understood and also due to the large amounts of data available which are recorded for regulatory purposes. Development of such ANN applications are made more difficult by the data handling issues. It is often advantageous to try techniques developed on different plant data sets, under different operating conditions etc, but this is difficult to achieve efficiently in software. In this research, however, we have described several different applications of ANNs to demonstrate the significant results which may be achieved. But more importantly we have shown how the ANN application development, modification, comprehensive testing and benchmarking can be efficiently completed. To simplify the presentation, we have omitted a section on neural network architectures and learning processes since this is adequately covered in the paper by Renotte et al included as part of this Special Feature.

Simulation and Pilot Plant Trials Aid Commissioning of Neuro Self-Tuning PI Controller

Abstract The paper presents the results of a simulation study and some pilot plant trials used in assessing the performance of a neuro self-tuning PI controller to be used (eventually) at a water treatment plant. The development and training of the adaptive loop is discussed and the benefits of neuro control over fixed gain PI control are demonstrated.