Research On Adaptive Control Of Temperature In Analyzer

In the industrial field, traditional PID algorithm is often used to control the temperature, location and flow, etc, and achieves good control result. However, with the continuous development of technology, some complex, changing systems appear in industrial control field, which limit the application of traditional control algorithm. At the same time, adaptive control theory has developed rapidly, and its typical theories have entered a more mature stage through years of research, such as fuzzy control theory, genetic algorithm theory and neural network theory. Due to the high specific and maintenance cost of these adaptive algorithms, the promotion and development of them in industrial control field is limited, therefore, a lot of improvement is definitely needed in practical application.The topic combines fuzzy theory and neural network theory with the traditional PID control, to gain fuzzy PID control and neural network PID control which are new algorithms suitable for industrial control application. The topic also summarizes the convergence, stability and anti-interference capacity of the new algorithms through resolving the practical temperature control problems of analyers'project platform, and compares various performances with traditional PID control to sum up the general selection standard of temperature control scheme in industrial analyers. Although many adaptive control theories have developed for a long time, they still stay in the simulation phase and lack practical application case. The topic puts the adaptive control theory into practical industrial products instead of simulating, which has been the main innovation.Fuzzy control theory is a logical description of control algorithm, with its core divided into: fuzzification interface, fuzzy rule base, fuzzy decision, and defuzzification. The integration process with the PID algorithm can be understood as follows: The PID parameters can be designed in the form of fuzzy rule base to dynamically change parameters in the convergence process in order to achieve better control results than PID. The experimental data show that the convergence and anti-interference capacity of fuzzy PID are better than PID. Therefore, for products which regular work in a harsh industrial environment, fuzzy PID control can be an alternative to traditional PID control. On the other hand, neural network system is a complex network system composed of interconnected neuron cells, which not only owns fast operation but also has strong adaptive learn- ing and analyzing ability. Its typical models are MP model and BP neural network model. The topic combines the MP model with incremental PID algorithm. And PID parameters work as the input weights of MP model, which can be amended using neural networkδlearning algorithm to achieve the purpose of temperature precise control.At the same time, there are some difficulties in the design process of fuzzy PID and neural network PID, such as: 1.the determination of the basic discourse and discrete discourse, 2.the formulation of fuzzy rule table, 3.the choice of membership function, 4.the choice of PID parameters'learning rateηin neural network PID, 5.the choice of point between PID and neural network PID. There is no uniform standard for those variables'choice, and the determination of those variables still base on engineers'experience, but with the continuous development of adaptive theory, there will be more and more application of adaptive theory in industry to cope with increasingly complex industrial control systems.