Fuzzy Network Deformation Measurement Research For Flexible Airfoil Long Baseline Antenna

The development of high-altitude long-endurance unmanned airborne early warning aircraft is receiving more and more attention, and the high-precision measurement in real time deformation of the airfoil long baseline antenna is of great significance to ensure the antenna performance. In this paper, the fuzzy theory is applied to the airfoil deformation measurement. Based on the universal approximation characteristic of the Takagi-Sugeno-Kang(TSK) fuzzy logic systems, two new fuzzy network methods which are independent of the measured object are proposed to approximate the relationship between the strain and displacement more accurately. The fuzzy network methods, which have the powerful capabilities for reasoning, adaptive learning and generality, provide a new way of thinking for the airfoil deformation measurement.For the reason that the anti-interference capability of the self-structuring fuzzy network(SSFN) is poor, a self-evolving interval type-2 fuzzy neural network(SSIT2FNN) is proposed through a combination of the neural network and interval type-2 fuzzy theory. In the SSIT2FNN, the antecedent part takes the type-2 fuzzy set model to form the feedback loop internally by feeding the acting strength of each rule, and it uses an algorithm of gradient-descent method for parameter learning. The consequent part takes the TSK model and uses a rule-ordered Karman filtering method for parameter learning. The initial rule number of the network is zero, and all rules are generated from the simultaneous on-line parameter learning from the antecedent part and consequent part.The SSIT2FNN is based on minimizing training error to train the network so that it relies on the training experience in the process of building a network to make the generalization performance of the system poor. Therefore, this paper combines the cluster splitting with the support vector regression(SVR) theory to propose a self-splitting iterative linear SVR fuzzy network(SSILSVRFN). The construction process of the SSILSVRFN is composed of structure and parameter learning. The structure learning uses a self-splitting rule generation(SSRG) algorithm for automatic rule number generation and initialization. For parameter learning, structural risk minimization by an iterative linear SVR(ILSVR) algorithm is proposed to iteratively optimize the antecedent and consequent part parameters. This paper demonstrates the capabilities of different methods by a simulation to approximate the same nonlinear function.Finally, an airfoil frame model has been designed, and then an experimental platform has been built for the deformation measurement experiment on the airfoil frame model under different kinds of static loads. The results show that the two kinds of improved fuzzy network algorithms have a high accuracy to approximate the relationship between the strain and displacement, and meanwhile, the deformation measurement method by fuzzy network is verified in actual effectiveness preliminarily.