Efthimios Kyriannakis

Model-based predictive control for generalized production planning problems

Abstract This paper introduces and elucidates the concept of model-based predictive control (MBPC) used as a decision-making tool for handling complex integrated production planning problems within a stochastic environment. First, a functional overview of the MBPC methodology is provided, then the framework under which the MBPC approach is used as a decision-making tool is given. Simulation experiments have strongly indicated the flexibility and effectiveness of the proposed approach.

Regulation of GMA welding thermal characteristics via a hierarchical MIMO predictive control scheme assuring stability

This paper deals with the regulation of the thermal characteristics of gas metal arc welding (GMAW). A complete treatment of the welding control problem requires the regulation of both the geometrical and thermal characteristics of the weld. Both classes of characteristics are of critical importance, but the thermal ones have received less attention in the majority of previous work. The present paper proposes a hierarchical predictive control scheme for the metallurgical characteristics of GMAW. A previously developed model for the regulation of the heat-affected zone, the cooling rate and the nugget cross-sectional area is used for the open-loop predictions. At the first level of the hierarchy, a parameterized generalized predictive control (GPC) algorithm is selected, among other control techniques, due to the inherited difficulty of the welding thermal process and the robustness of this algorithm against modeling errors and parameter variations. The main drawback of GPC is that it does not guarantee stability. The second level of the control hierarchy, where a coordinator specifies a set of reliable values for the parameters of GPC, so that stability is assured, and transmits them to the controller of the first level overcomes this difficulty. This paper provides a representative set of simulation results obtained by the present hierarchical GPC scheme, including a comparison with the classical linear-quadratic optimal control scheme. These results show the superiority of the GPC scheme.

Geometry and Thermal Regulation of GMA Welding via Conventional and Neural Adaptive Control

This paper investigates the application of conventional and neural adaptive control schemes to Gas Metal Arc (GMA) welding. The goal is to produce welds of high quality and strength. This can be achieved through proper on-line control of the geometrical and thermal characteristics of the process. The welding process is variant in time and strongly nonlinear, and is subject to many defects due to improper regulation of parameters like arc voltage and current, or travel speed of the torch. Adaptive control is thus naturally a very good candidate for the regulation of the geometrical and thermal characteristics of the welding process. Here four adaptive control techniques are reviewed and tested, namely: model reference adaptive control (MRAC), pseudogradient adaptive control (PAC), multivariable self-tuning adaptive control (STC), and neural adaptive control (NAC). Extensive numerical results are provided, together with a discussion of the relative merits and limitations of the above techniques.

ON–LINE IMPROVEMENT FOR THE DECENTRALIZED PREDICTIVE CONTROL OF THE HEAT DYNAMICS OF A GREENHOUSE

Abstract A decentralized model-based predictive controller is used for the design of discrete-time control systems aiming at regulating the air temperature and heat supply in greenhouses. Moreover, alternative techniques are proposed for the approximation of the decentralized part of the control and the on-line improvement of the overall control problem. A state space model is used to predict the corresponding local indoor temperature over a long-range time-period and approximate models are used to predict the interactions among the subsystems. The sun radiation and outdoor temperature are treated as external disturbances that affect the overall system dynamics. A series of energy fluxes consist the heating system and the predictive controllers have proved to be powerful in controlling the supply temperature.

Pseudo-Derivative Feedback Based Identification of Unstable Processes with Application to a Bioreactor

Abstract The use of a pseudo-derivative feedback structure for parameter identification of unstable first order plus dead-time processes is investigated. Two alternative identification methods are presented. The proposed identification methods require small computational effort and they are particularly useful for on-line applications. They ensure considerably smoother response (i.e. less overshoot), than known identification methods based on standard PI/PID controllers, while providing better accuracy and more simplicity in implementation, due to the simpler structure of the pseudo-derivative feedback controller used.

A Fuzzy Adaptive - Predictive Scheme for the Decentralized Control of a Class of Interconnected Systems

Abstract This paper proposes an approach for the design of discrete-time decentralized control systems with m-step delay sharing information pattern, employing the modelbased predictive control (MBPC) scheme combined with fuzzy prediction for the interconnections among the subsystems. A state-space model is used at each control station to predict the corresponding subsystem output over a long-range time period. The interaction trajectories are considered to be non-linear functions of the states of the subsystems. For all cases the interconnections and the necessary predictions for them are estimated by an appropriate adaptive fuzzy identifier based on the generation of linguistic IF-THEN rules and the on-line construction of a common fuzzy rule base. Representative computer simulation results are provided and compared for nontrivial example systems.

Decentralized Predictive Control of Large Scale Systems Using Neuro-Fuzzy Identifiers for their Interconnections

Abstract This paper proposes an approach for the design of discrete-time decentralized control systems covering not only the case of m-step delay sharing information pattern, but also any general nonclassical information pattern where the non-local information is not spread among the subsystems. It employs the model-based predictive control (MBPC) scheme combined with fuzzy prediction for the interconnections among the subsystems. A state space model is used at each control station to predict the corresponding subsystem output over a long-range time period. The interaction trajectories are considered to be non-linear functions of the states of the subsystems. In all cases, the interconnections and the necessary predictions for them are estimated by an appropriate neuro-fuzzy identifier trained on-line using the back-propagation training algorithm. Representative computer simulation results are provided and compared for nontrivial example systems.

GMA Welding Process Regulation via Hierarchical GPC with Assured Stability

Abstract This paper deals with the regulation of the thermal characteristics of Gas Metal Arc (GMA) welding. A previously developed model is used for the open loop predictions. At the first level of the hierarchy, a parameterized Generalized Predictive Control (GPC) algorithm is selected among other control techniques, due to its robustness against modeling errors and parameter variations. A coordinator, at the second level of the hierarchy, specifies a set of reliable values for the parameters of GPC, so that stability is assured. A representative set of simulation results is included, along an evaluation of the advantages and limitations of the proposed hierarchical GPC technique.

Regulation of GMA Welding Process Features Using Neural Adaptive Control

Abstract Neural adaptive control (NAC) is naturally a very good candidate for the regulation of the geometrical and thermal characteristics of the welding process which is subject to strong variations and disturbances. Here two different versions of neural adaptive control are tested. The applied control scheme is the Feedback Error Lerning Architecture (FELA). In its flrst version a Multilayered Perceptron (MLP) Neurocontroller is used while its second version is based on a Radial Basis Function (RBF) Neurocontroller. Extensive numerical results are provided along with a discussion of the relative merits and limitations of the above techniques.

Neural Fault Detection and Accommodation for Interconnected Power Systems

Abstract Due to increasing safety and reliability demands fault detection and accommodation is a key issue in the design of industrial control systems. This paper presents a neural network based method for fault detection and accommodation, appropriate for interconnected power systems. The simulation results show that the method is very promising.