SUGENO-TYPE FUZZY NEURAL NETWORK ACTIVE CONTROL OF SPACE FRAME STRUCTURE BASED ON PIEZOELECTRIC ACTUATOR

Based on the vibration control characteristics of a piezoelectric active-member invented independently, a Sugeno-type fuzzy neural network control system of a space frame structure is designed, in which the inputs are the relative displacement and relative speed of the two nodes at the end of the active-member and the output is the control current. First, the LQR method is used to obtain the training data samples by controlling the structure, then the adaptive learning function of neural network is used to do fuzzy partition and generate fuzzy rules, and at last a space frame structure model is actively controlled by using fuzzy reasoning capability under the action of seismic response, where the result is compared with the result produced by the simulation of Mamdani fuzzy inference rules based on experiences. The results show that both two fuzzy reasoning models achieve good control effects, but the simulation speed of the Sugeno fuzzy inference is dozens of times faster than the Mamdani fuzzy inference because of its simple calculation that disregards the human experiences, thus it can meet the application requirements better by using the Sugeno-type fuzzy neural network controller.