Research On FPGA-Based Fuzzy Predictive Control Of Grain Drying Process

As grain drying is a complex, time-dependent, nonlinear and long delay industrial process, it's very difficult to implement control process. And the research of grain drying process control technology, control model, process realization and application become hot.China is a big production country of grain. Then it has an important meaning to independently develop medium-sized or pint-sized grain dryers that suit medium-sized farmers or dirt farmers and realize process automatic control. It is important to insure the grain moisture in well uniformity, to improve drying grain quality, greatly to reduce grain waste, to reduce the labor intensity of operations and at last to make full use of the dryer capacity.Research is based on fuzzy predictive control strategy and FPGA technology to design and realize the controller. With the advantages of large-scale, high level of integration, high reliability and short development cycle, convenient and flexible and with the characteristics of low cost, low power consumption and high reconfiguration, field-programmable technology provides a very good solution for the controller development. The main contents of this paper are as follows.(1) Designing fuzzy predictive controller for grain drying process with the combination of fuzzy control and predictive control. Predictive control is suitable for grain process which is time-dependent, nonlinear, long delay. Fuzzy control can be applied to control process by using rich experience and to improve the control's accuracy.(2) Building the predictive model of the controller based on the drying principle of the establishment of deep bed, and modifying model with feedback correction; take the error and error changes between model predictive value and set point value as fuzzy control inputs; building two-dimensional fuzzy control and the controller result is gained after the steps of fuzzed, fuzzy reasoning and defuzzy.(3) Researching on float operation units based on FPGA. As predictive model needs float operations, the paper normalizes data as the format of one bit mantissa symbol, 11 bits mantissa, one bit index symbol and 8 bits inde~x. And then analyze, design and realize FPGA-based float addition algorithm, float multiplication algorithm, float division algorithm and e~x float algorithm. Float addition algorithm first aligns the inde~x and then adds the data. Float multiplication algorithm uses displacement operations, iterative operations and addition to achieve the results. Float division uses Goldschmidt-based iterative algorithm to calculate results. The integer part of e~x uses look-up table to find the results and the pure decimal part of e~x uses CORDIC algorithm. The result is obtained from multiplying results of those two parts.(4) Realization of fuzzy predictive controller based on FPGA. This part designs a limited state machine based on VHDL. The state machine, all the basic float data computing units and the fuzzy control unit constitute a whole fuzzy predictive controller.