James H. Taylor

An Expert System Architecture for Coping with Complexity in Computer-Aided Control Engineering

Abstract We discuss an architecture for an expert system to assist a control engineer in coping with the complexity of computer-aided control engineering, with major emphasis placed on implementation. Issues to be treated include: rule base organization, knowledge to be contained in various rule bases, a mechanism to switch between rule bases as the problem solution proceeds, a protocol to coordinate the symbolic computations of the inference engine with the numeric computations of conventional analysis and design software, support of non-monotonic reasoning to permit retracting and revising steps in the design process, and conversion between numeric and symbolic data. We present our approach to these issues, and give examples of rules used to represent operational knowledge and to guide the solution of the problem.

Nonlinear Compensator Synthesis via Sinusoidal-Input Describing Functions

We describe a new nonlinear compensator synthesis approach and illustrate it with an application to a position servo design problem from robotics. The synthesis technique is based on a set of amplitude-dependent sinusoidal-input describing function (SIDF) models of the nonlinear plant. An intermediate step is the design of a linear compensator set based on these models; final synthesis of the nonlinear control system is accomplished by SIDE inversion to determine the required compensator nonlinearities. The major extension in comparison with earlier research is that the compensator so obtained is fully nonlinear; i.e., there is a nonlinear operator associated with each term (proportional, integral, derivative) in the compensator. This approach is capable of treating nonlinear plants of a very general type, with no restrictions as to system order, number of nonlinearities, configuration, or nonlinearity type, and can be extended readily to include other compensator types, e.g., lead/lag. The end result is a closed-loop nonlinear control system that is relatively insensitive to reference input amplitude.

Applications of a Nonlinear Controller Design Approach based on Quasilinear System Models

In this paper, we report on recent progress in developing nonlinear control system design techniques based on sinusoidal-input describing function (SIDF) methods. Primarily, this involves illustrating a fundamental difference between SIDF and random-input describing function (RIDF) models of nonlinear systems, developing the nonlinear controller design method more fully, and demonstrating it by applying it to a significant nonlinear control design problem in robotics. Based on these results, the use of this nonlinear controller design method should be substantially better understood.

Prototype design of a multi-agent system for integrated control and asset management of petroleum production facilities

This paper addresses a practical intelligent multi- agent system for asset management for the petroleum industry, which is crucial for profitable oil and gas facilities operations and maintenance. A research project was initiated to study the feasibility of an intelligent asset management system. Having proposed a conceptual model, architecture, and implementation plan for such a system in previous work and defined its autonomy, communications, and artificial intelligence (AI) requirements, we are proceeding to build a system prototype and simulate it in real time to validate its logical behavior in normal and abnormal process situations. We also conducted a thorough system performance analysis to detect any computational bottlenecks. Although the preliminary system prototype design has limitations, simulation results have demonstrated an effective system logical behavior and performance.

Expert-aided Control Engineering Environment for Nonlinear Systems

Abstract We have been working to develop an expert-system-based environment for Computer-Aided Control engineering (CACE). Our goal is to create a high-level user interface to conventional CACE tools, with substantial capabilities in areas that are either very complicated, or that require heuristic logic or specialized knowledge, or both. So far, we have developed rule bases for linear system diagnosis, specification development, lead lag compensator synthesis, and design validation. We report here on recent research in expert-aided CAGE for nonlinear systems. Our approach is to perform extensive experimentation (using both numerical and symbolic processing) under the direction of a rule base containing “expertise” in and heuristic strategies for nonlinear CACE. The results and status of this effort are described in detail. This work represents the first phase in an iterative process; much is being learned that will be folded back into the rule base to improve the capabilities of the expert system. We should stress that our expert system is not particularly deep at this time - our present objective is to aid the user in making the most effective use of rather complicated procedures and conventional CACE software; in many instances, this simply involves “common sense”.

A Nonlinear PID Autotuning Algorithm

A nonlinear autotuning regulator algorithm is obtained via a direct combination of the Åström-Hägglund algorithm for the linear case [1] with the sinusoidal-input describing function (SIDEF) approach to nonlinear compensator synthesis of Taylor and Strobel [2]. The basic approach for linear autotuning proceeds as follows: a. install a relay with hysteresis in series with the unknown plant to be controlled; close a unitygain feedback loop around this combination; b. choose several values of hysteresis so that this system exhibits limit cycles; the frequencies and amplitudes of the oscillation at the output of the plant determine points on the plant Nyquist plot; and c. given points on the plant Nyquist plot, set the PID controller gains using an appropriate tuning algorithm (e.g., Ziegler-Nichols). This approach produces good results if the plant is liner or nearly so; however, if the plant behavior is strongly amplitude-dependent, there are likely to be problems with implementing this algorithm. The nonlinear autotuning regulator algorithm which extends the above approach to handle situations where the plant behavior is strongly amplitude-dependent is based on the SIDF approach. In essence, SIDF input/output (I/O) models of the compensated nonlinear system are exploited to directly synthesize a compensator nonlinearity that eliminates or reduces the amplitude dependence of the open-loop I/O relation. The nonlinear synthesis portion of this algorithm is reasonably simple to implement, has been shown to be effective [2], and should be of practical utility. An example application to a precision position control system is provided as an illustration.

Rigorous hybrid systems simulation of an electro-mechanical pointing system with discrete-time control

Earlier research in the modeling and simulation of hybrid systems led to the development of a general hybrid systems modeling language (HSML). Effort is underway to implement this concept in software. The standard MATLAB model framework and integration algorithms have been extended to support state-event handling in continuous-time components and to handle embedded discrete-time components. In this paper we overview the algorithmic implementation of the HSML language constructs for dealing with state events and embedded discrete-time modules in MATLAB. An extensive example of an electro-mechanical pointing system with stiction under discrete-time control is presented to demonstrate the efficacy of these extensions.

Fault Detection and Isolation Using the Generalized Parity Vector Technique in the Absence of an a Priori Mathematical Model

This paper is an extension of the generalized parity vector (GPV) approach presented in Omana and Taylor (2005, 2006). In the present work, this fault detection and isolation (FDI) technique is implemented on a two-phase separator followed by a three-phase gravity separator model used in oil production facilities. This model simulates a larger scale process, which allows the technique to be tested in a higher dimensional space with more complex system dynamics. Also, the plant model availability issue is overcome by incorporating a system identification module before executing the FDI block. This shows that while the GPV is a model-based technique, it is still viable for FDI even for those plants where only input-output data are available.

Future Developments in Modern Environments for CADCS

Abstract Recent and future efforts at GB to develop modem environments for CADCS are discussed. The basic elements of these systems are: • a User Interface which combines a “point-and-click” menu-and forms-driven interface with other access modes for the more experienced user • a Data-Base Manager organized in terms of Projects, Models and corresponding Results and other related data elements and including version control • an Expert System Shell, which performs routine higher-level CACE tasks, and • a data-driven Supervisor that integrates the above elements with existing CACE packages for linear and nonlinear simulation, analysis and design As is usually the case, it has been leamed that much more can be done to provide a fully supportive environment for controls engineering, and it has also become clear that certain things might better be done differently. This presentation will focus on such areas, especially on generic issues that can be applied to other CADCS systems

Ehanced sensor/actuator resolution and robustness analysis for FDI using the extended generalized parity vector technique

This paper is an extension of the generalized parity vector (GPV) approach presented by the authors (2005), Some aspects of sensor isolation are first clarified and a special case is defined to solve an important sensor/actuator fault detection and isolation (FDI) ambiguity issue. To overcome this problem, a new optimization constraint is incorporated in the transformation generation procedure used to improve separation in the generalized parity space. The validity of the different aspects analyzed through this research is demonstrated by testing this FDI scheme using a nonlinear jacketed continuously stirred tank reactor model. Robustness analysis is then performed over the controller envelope, showing the capability of the FDI technique to handle operating point and fault size variability