Structural Fuzzy Reliability Analysis Based On Extreme Learning Machine Multiple Response Surface Method

The complexity of mechanical structure is getting higher and higher,the design requirements are gradually improved,and the reliability analysis of complex structures is getting more and more attention.The aero-engine is the propulsion system of the aircraft.The failure of the blade is always a serious problem in the engine,so it is of great practical significance to accurately analyze the blade.In order to improve the accuracy and efficiency of reliability analysis,this paper combines the nonlinear fitting ability of the extreme learning machine neural network with the response surface method and applies it to the blisk model for verification.The main contents include:(1)Blade reliability analysis based on the ultimate learning machine response surface method.The finite element model of three-dimensional blade is established.The material density,elastic modulus,rotational speed and aerodynamic force of the blade are selected as random input variables.The deformation of the blade is the output response.The static mechanical analysis of the blade is performed by ANSYS Workbench software to obtain the contact deformation of the blade.The distributed cloud map is extracted by using the Latin hypercube sampling technique to obtain 100 sets of random variables,and the corresponding 100 sets of responses are obtained.The limit learning machine response surface method is used to train the sample to obtain the response machine function of the extreme learning machine,and Monte Carlo method is used.The function is sampled 10,000 times,and the random deformation value obtained by sampling is compared with the allowable deformation value to obtain the reliability.(2)Blisk reliability analysis based on multiple learning surface method of extreme learning machine.When the finite element simulation of the leaf disc is carried out,considering the influencing factors of the leaf disc under the flow-solid coupling field,the material density,rotation speed and inlet flow velocity of the leaf disc are random input variables,and the deformation,stress and strain of the leaf disc are output responses.Through the multiple learning surface method of the extreme learning machine to establish the functional relationship between the blisk output response and the random variable,the explicit function expression is obtained.Considering the correlation of the leaf disk failure mode,the Monte Carlo method is used to link the equations.Sampling,and then calculating the reliability of the blisk.The comparison of the methods shows that the multiple learning surface method of the extreme learning machine has high calculation accuracy and calculation speed.(3)Anti-resonance fuzzy reliability analysis of blade based on response learning surface method of extreme learning machine.In order to accurately calculate the anti-resonance reliability of turbine blades,this paper selects the material density,blade height,tip thickness,rotational speed and aerodynamic pressure of the blade as random input variables.The natural frequency of the blade is the output response,and the modal analysis is performed on the blade.The function model of the relationship between random input variables and response is obtained by using the extreme learning machine response surface method,and the function model is sampled by MCM method to obtain the first 6 random natural frequency values under typical speed.Drawing the Campbell diagram can intuitively find the intersection of the dynamic frequency curve and the excitation force frequency curve,which is the resonance point that may occur,and the corresponding rotational speed is the resonance speed.Since the probability of the blade not resonating is that the operating speed of the blade avoids the resonance speed,considering the ambiguity of the resonance state,the resonance reliability of the blade is obtained from the blade anti-resonance fuzzy reliability model.(4)Fuzzy Reliability Analysis of blisk Vibration Based on Extreme Extremum Response Surface Method of Extreme Learning Machine.Based on the consideration of the randomness of design variables and the ambiguity of ambiguity and intermediate transition state,the fuzzy reliability analysis of blisk vibration based on the extreme extremum response surface method of extreme learning machine is proposed.The transformation method converts the fuzzy variable into an equivalent random variable,so that the reliability problem with fuzzy variables is transformed into a problem controlled by random variables,and the reliability of the blisk vibration is obtained by using the fuzzy reliability calculation model.The calculated reliability results can better reflect the actual situation of the blisk.