No abstract

Real control systems require robust control performance to deal with unpredictable and altering operating conditions of real-world systems. Improvement of disturbance rejection control performance should be considered as one of the essential control objectives in practical control system design tasks. This study presents a multi-loop Model Reference Adaptive Control (MRAC) scheme that leverages a nonlinear autoregressive neural network with external inputs (NARX) model in as the reference model. Authors observed that the performance of multi-loop MRAC-fractional-order proportional integral derivative (FOPID) control with MIT rule largely depends on the capability of the reference model to represent leading closed-loop dynamics of the experimental ML system. As such, the NARX model is used to represent disturbance-free dynamical behavior of PID control loop. It is remarkable that the obtained reference model is independent of the tuning of other control loops in the control system. The multi-loop MRAC-FOPID control structure detects impacts of disturbance incidents on control performance of the closed-loop FOPID control system and adapts the response of the FOPID control system to reduce the negative effects of the additive input disturbance. This multi-loop control structure deploys two specialized control loops: an inner loop, which is the closed-loop FOPID control system for stability and set-point control, and an outer loop, which involves a NARX reference model and an MIT rule to increase the adaptation ability of the system. Thus, the two-loop MRAC structure allows improvement of disturbance rejection performance without deteriorating precise set-point control and stability characteristics of the FOPID control loop. This is an important benefit of this control structure. To demonstrate disturbance rejection performance improvements of the proposed multi-loop MRAC-FOPID control with NARX model, an experimental study is conducted for disturbance rejection control of magnetic levitation test setup in the laboratory. Simulation and experimental results indicate an improvement of disturbance rejection performance.

Neuroevolutionary machine learning is an emerging topic in the evolutionary computation field and enables practical modeling solutions for data-driven engineering applications. Contributions of this study to the neuroevolutionary machine learning area are twofold: firstly, this study presents an evolutionary field theorem of search agents and suggests an algorithm for Evolutionary Field Optimization with Geometric Strategies (EFO-GS) on the basis of the evolutionary field theorem. The proposed EFO-GS algorithm benefits from a field-adapted differential crossover mechanism, a field-aware metamutation process to improve the evolutionary search quality. Secondly, the multiplicative neuron model is modified to develop Power-Weighted Multiplicative (PWM) neural models. The modified PWM neuron model involves the power-weighted multiplicative units similar to dendritic branches of biological neurons, and this neuron model can better represent polynomial nonlinearity and they can operate in the real-valued neuron mode, complex-valued neuron mode, and the mixed-mode. In this study, the EFO-GS algorithm is used for the training of the PWM neuron models to perform an efficient neuroevolutionary computation. Authors implement the proposed PWM neural processing with the EFO-GS in an electronic nose application to accurately estimate Nitrogen Oxides (NOx) pollutant concentrations from low-cost multi-sensor array measurements and demonstrate improvements in estimation performance.

Identification and prediction of clogging behavior in heating, ventilation, and air conditioning (HVAC) filters is crucial to avoid issues such as system overheating, energy waste, lower indoor air quality, etc. Researchers are focusing more on the particle loading characteristics of a filter medium in a laboratory environment under steady-state conditions, fixed particle concentrations, area of porosity, dust feed and volumetric flow rate. However, recent research still shows uncertainties in modeling as well as the implementation problems of constructing the HVAC laboratory test bench and equipment. In addition, subjects such as non-uniform particle deposition depreciation of the condition and various type of mechanical filters such as fibrous, fabric, granular, and membrane filter or electrostatic filters which typically used in HVAC systems perform under some assumptions and still need more research. The studies become even more difficult acquiring a large number of time-varying and noisy signals. Another approach among studies is data-driven knowing that Building Automation System (BAS) is not equipped with appropriate sensor measuring the clogging, it is needed to drive the clogging mathematical model from the pressure drop signal. This paper bridges the gap between particle-size study and black box modeling of HVAC filter which has not received much attention from authors. The proposed method assumes that the pressure drop is the result of two time-varying functions; f(t), which represents the dynamics of clogging and, g(t), which refers to dynamics of remained terms. The exponential and polynomial of second order functions are proposed to express the clogging behavior. The software package based on Particle Swarm Optimization Artificial Bee Colony (PSOABC) algorithm, is developed and implemented to estimate the coefficients of the clogging functions based on smallest RMSE, high coefficient of correlation and acceptable tracking. Five Air Handling Unit (AHUs) are selected for practical verification of the model and the results show that the applied method can successfully predict clogging and pressure drop behaviour of HVAC filters.