Accurate Plant Model Research Articles

Improvement of the convergence property of adaptive feedforward controllers and their application to the active control of ship interior noise

Various structural measures against vibration and noise were taken in a training ship, the Oshima Maru. However, an unpleasant sound persisted in the mess hall, where crews take their breaks. To reduce the noise, active controllers were investigated to satisfy the causality constraint in their update. Some of the controllers were preconditioned using the inverse of the plant because their convergence rates are limited by the dynamics and coupling within the plant response. The overall response from the output of the control filter to the output of the error sensor is thus equal to the all-pass part of the plant response. Because this response has a flat magnitude, the convergence speed of the adaptive algorithm is not affected by resonances in the plant response, as it is for the normal filtered reference LMS algorithm. The algorithms were compared under the same conditions to investigate differences in their properties and also corrected to satisfy the causality of their update processes. A comparison of convergence properties shows that satisfying the causality constraint in their update results in an improvement, and consequently the effect was confirmed by the filtered reference - filtered error LMS algorithm. Simulations are presented for a control system that was introduced using plant responses measured from a loudspeaker to a microphone in the mess hall inside the Oshima Maru. After investigating the convergence speed in various gradient descent adaptation algorithms, the results were integrated with the actual plant response and applied to the active control of ship interior noise. It was also found that although the preconditioned LMS algorithm converges dramatically faster than the ordinary gradient descent adaptation algorithms with an accurate plant model, its convergence rate is still sensitive to the auto-correlation and cross-correlation properties of the reference signals.

COMPARISON OF ADAPTIVE CONTROLLERS FOR MULTICHANNEL FEEDFORWARD CONTROL AND THEIR APPLICATION TO THE ACTIVE CONTROL OF SHIP INTERIOR NOISE

Various structural measures against vibration and noise were taken in a training ship, Oshima Maru. However, an unpleasant sound persisted in the mess hall, where crews take their breaks. In order to reduce the noise, active controllers were investigated. Some of them were preconditioned using the inverse of the plant because their convergence rates are limited by the dynamics and coupling within the plant response. The algorithms were compared under the same conditions to investigate differences in their properties and also corrected to satisfy the causality of their update processes. Simulations for a control system were introduced using plant responses measured from a loudspeaker to a microphone in the mess hall inside Oshima Maru. After investigating the convergence speed in various gradient descent adaptation algorithms, the results were integrated with the actual plant response and applied to the active control of ship interior noise. It was also shown that the although preconditioned LMS algorithm converges dramatically faster than the ordinary gradient descent adaptation algorithms with an accurate plant model, its convergence rate is still sensitive to the autocorrelation and cross-correlation properties of the reference signals.

Trajectory tracking of a batch polymerization reactor based on input–output-linearization of a neural process model

Input–output-linearization via state feedback offers the potential to serve as a practical and systematic design methodology for nonlinear control systems. Nevertheless, its widespread use is delayed due to the fact that developing an accurate plant model based on physical principles is often too costly and time consuming. Data-based modeling of dynamic systems using neural networks offers a cost-effective alternative. This work describes the methodology of input–output-linearization using neural process models and gives an extended simulative case study of its application to trajectory tracking of a batch polymerization reactor.

Fault detection and accommodation in dynamic systems using adaptive neurofuzzy systems

Fault detection and accommodation plays a very important role in critical applications. A new software redundancy approach based on an adaptive neurofuzzy system (ANFIS) is introduced. An ANFIS model is used to detect the fault while another model is used to accommodate it. An accurate plant model is assumed with arbitrary additive faults. The two models are trained online using a gradient-based approach. The accommodation mechanism is based on matching the output of the plant with the output of a reference model. Furthermore, the accommodation mechanism does not assume a special type of system or nonlinearity. Simulation studies prove the effectiveness of the new system even when a severe failure occurs. Robustness to noise and inaccuracies in the plant model are also demonstrated.

ACTIVE POSITION CONTROL OF A FLEXIBLE SMART BEAM USING INTERNAL MODEL CONTROL

The problem of controlling the position at the tip of a flexible cantilever beam to follow a command signal is considered, by using a pair of piezoelectric actuators at the clamped end. The beam is lightly damped and so the natural transient response is rather long, and also since the sensor and actuator are not collocated, the plant response is non-minimum phase. Two control strategies were investigated. The first involved conventional PID control in which the feedback gains were adjusted to give the fastest closed-loop response to a step input. The second control strategy was based on an internal model control (IMC) architecture. The control filter in the IMC controller was a digital FIR device designed to minimize the expectation of the mean square tracking error. In practice, such smart beams could be exposed to temperature fluctuations and changes in geometry. The effect of these variations on the stability was studied and it is shown that the need for robustness to such variations leads to a limitation in the performance of an IMC controller. The improvement in the stability robustness by incorporating control effort weighting into the cost function being minimized was investigated, as was the incorporation of modelling delay in the design of the IMC control filter. The IMC controller designed for the beam was found to have much reduced settling times to a step input compared with those of the PID controller while maintaining good robustness to changes in temperature. However, the extremely low damping of the experimental beam made it difficult to implement an accurate plant model in practice.

Input/output linearization using time delay control and time delay observer

An input/output linearization (IOL) method using time delay control (TDC) and time delay observer (TDO) is presented. This method enables the IOL method to be applied to plants when all the states of plant are not measurable or the measured plant output is very noisy. The designed control system requires neither an accurate plant model nor the real time computation of plant nonlinearity. Consequently, the proposed control algorithm turned out to be computationally efficient and easy to design for nonlinear plants. In a simulation for a second order nonlinear plant, the output followed desired response well and the control performance appeared to be superior to IOL using TDC and numerical differentiation. Finally in an experiment with a pneumatic servo system, we obtained results consistent with those obtained from the simulation, and it was confirmed that the proposed control algorithm can be effectively used in a real closed-loop system.

Modeling of an Acoustic Path and PID Feedback Control Systems Design for Active Noise Control.

Focus is placed on the feedback (FB) active noise control (ANC) in order to reduce costs by simplifying ANC systems, because FB-ANC does not need reference signals. The purpose of this paper is to obtain a low order accurate model of an acoustic plant for FB-ANC and design a simple controller based on this model. The plant in this case is from the input to a single secondary loudspeaker to the output of a single error sensor (microphone). First, as a modeling method, physical modeling is applied. Second, Smith predictor is modified for FB-ANC to eliminate the influence of time delay in an acoustic field, and a PID controller is designed. Finally, effectiveness of the modeling methods and the proposed FB-ANC are demonstrated through numerical experiments using actual data.

The application of neural nets in the metallurgical industry

Although the potential of new techniques for the construction of accurate plant models, such as those based on connectionist methods, is generally acknowledged, little on their practical application can be found in the chemical and metallurgical engineering literature. In this paper the use of neural nets to model gold losses on a reduction plant and the consumption of an additive on a leach plant, as well as the pyrometallurgical processing of zinc and aluminium is discussed. The gold and leach plant models performed better than the multilinear regression models used on the plants, even where relatively few data were available. The neural networks used to model the recovery of lead and zinc from industrial flue dusts, process synthesis of zinc recovery plants and the processing of secondary aluminium in a rotary salt flux furnace produced realistic results that could be used by plant personnel to optimize their operations.