基于PSO优化小波神经网络的无人机动力系统故障诊断模型

UAV power based on PSO optimized wavelet neural network system fault diagnosis model

针对传统小波神经网络对无人机动力系统的故障信号降噪和识别能力差以及网络收敛速度慢、训练精度不高的问题,构建了基于改进粒子群算法(PSO)优化小波神经网络的无人机动力系统故障诊断模型。该模型运用软、硬阈值函数组合改进的新阈值函数和改进PSO优化小波神经网络的方式,克服重构信号不连续或严重失真的问题,优化了小波神经网络初始权值和阈值,使模型能够实现快速、准确分析和识别故障类型,具有较好的故障预测和诊断能力。本文中通过对比不同阈值函数的降噪能力和PSO、GA、ACO对小波神经网络的改进效果,比较BP神经网络、传统小波神经网络、PSO优化小波神经网络的故障诊断预测效果,验证了本文中构建的PSO优化小波神经网络故障诊断模型远优于其他对比模型,具有故障识别和降噪能力强、收敛速度快、训练精度高的优点,在无人机动力系统的故障诊断领域,具有较好的可行性和有效性。

Aiming at the problems of poor noise reduction and recognition ability of traditional wavelet neural network for fault signals of UAV power system and slow convergence speed and low training accuracy of the network,a fault diagnosis model of UAV power system based on improved particle swarm algorithm(PSO)optimized wavelet neural network is constructed.The model uses a combination of soft and hard threshold functions improved new threshold functions and improved PSO optimized wavelet neural network to overcome the problem of reconstructed signal discontinuity or severe distortion and optimize the initial weights and thresholds of the wavelet neural network,so that the model can achieve fast and accurate analysis and identification of fault types with better fault prediction and diagnosis capability.In this paper,by comparing the noise reduction ability of different threshold functions and the improvement effect of PSO,GA and ACO on wavelet neural network,and comparing the fault diagnosis prediction effect of BP neural network,traditional wavelet neural network and PSO optimized wavelet neural network,it is verified that the fault diagnosis model of PSO optimized wavelet neural network constructed in this paper is much better than other comparison models,with strong fault identification and noise reduction ability.It has the advantages of high fault identification and noise reduction ability,fast convergence speed and high training accuracy,and has good feasibility and effectiveness in the field of fault diagnosis of UAV power system.