Design of intelligent control system and its application on fabricated conveyor belt grain dryer
The grain drying process is one of the most critical post-harvest operations in modern agricultural production. Efficient control of this process is an indispensable requirement especially in light of recent demands for handling latest increase in energy costs,achieving current requirement for eco-f...
Saved in:
Main Author: | |
---|---|
Format: | Thesis |
Language: | English |
Published: |
2011
|
Online Access: | http://psasir.upm.edu.my/id/eprint/42184/1/FK%202011%2058R.pdf http://psasir.upm.edu.my/id/eprint/42184/ |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | The grain drying process is one of the most critical post-harvest operations in modern agricultural production. Efficient control of this process is an indispensable requirement especially in light of recent demands for handling latest increase in energy costs,achieving current requirement for eco-friendly technologies, and producing products of high quality.The grain drying process is characterized by its complex nature. As a result, the mathematical models developed for these systems consist of sets of highly complex and nonlinear partial differential equations (PDEs) which require highly complicated numerical techniques to solve them. Therefore, these models are not suitable for control system design. Moreover, despite the complexity of the drying process, grain dryers, in particular conveyor-belt grain dryers, are still controlled by conventional PID controllers. The major objective of this research is to improve the performance of conveyor-belt grain dryers by designing an intelligent control system utilizing the capabilities of the adaptive neuro-fuzzy inference system (ANFIS) to model and control the drying process. To achieve this objective, a laboratory-scale conveyor-belt grain dryer was specifically fabricated for this study. As the main controller in this work, a simplified ANFIS structure is proposed to act as a proportional-integral-derivative (PID)-like feedback controller to control nonlinear systems. This controller has several advantages over its conventional ANFIS counterpart, particularly the reduction in processing time. Moreover, three evolutionary algorithms (EAs), in particular a real-coded genetic algorithm (GA), a particle swarm optimization (PSO), and a global-best harmony search (GHS), were separately used to train the proposed controller and to determine its scaling factors. These EAs overcome a common problem encountered in derivative-based learning methods, which is the necessity for the teaching signal when applying the ANIFIS as a controller.To demonstrate the effectiveness of the proposed controller, several non-linear plant models were used to evaluate its performance in terms of control accuracy, generalization ability, and robustness against external disturbances and parameter variations in the controlled system. In addition, several comparative studies were conducted with other related controllers, namely a conventional ANFIS controller, a variation of the ANFIS network called complex fuzzy basis function network (CFBFN),and a conventional PID controller. Furthermore, the ability of the simplified ANFIS controller to control nonlinear multi-input multi-output (MIMO) systems was also investigated. The results of all these tests clearly indicated the notable performance of the proposed controller. After fabricating the conveyor-belt grain dryer, a real-time experiment was conducted to dry paddy grains, in particular the MR 219 rice variety. The grains were first re-wetted to a moisture content (MC) of about 18% wet basis (wb). Next, by fixing the dryer operating conditions of temperature, flow rate, and humidity of the drying air, the voltage to the dryer motor was manipulated in a pre-specified sequence to give the required conveyor-belt speed for each paddy sample. The corresponding MC of each of these samples was measured by the XM 120 Moisture Analyzer. The result was a set of 50 input-output samples which were then presented to an ANFIS network to develop the desired process model. The modeling performance achieved by this ANFIS model was then compared with those of an autoregressive with exogenous input (ARX) model and an artificial neural network (ANN) model, and the results clearly showed the superiority of the developed ANFIS model. The simplified ANFIS controller was then applied to control the developed ANFIS-based dryer model using different initial conditions based on real data. In addition, five robustness tests were made to evaluate the controller ability in handling unexpected changes in the drying operating conditions. Furthermore, a comparative study with a genetically-tuned PID controller was conducted. From all these tests, the simplified ANFIS controller has proved its remarkable ability in controlling the grain drying process represented by the developed ANFIS model. |
---|