基于改进型DBN深度学习的复合黏接质量模式识别研究

Study on Pattern Recognition of Composite Bonding Quality Based on Improved DBN Deep Learning

针对薄板复合材料粘接质量识别问题,提出了一种基于粒子群算法(PSO)的改进型深度置信网络(DBN)深度学习模型,使用PSO来优化标准DBN网络权重,以解决薄板复合材料在识别过程中分类过拟合的问题,并利用该模型完成了薄板复合材料粘接质量的识别过程。探讨算法设计步骤,并对模型进行多次调试、训练,确定网络模型参数,完成测试数据识别。仿真结果表明:经过多次训练测试调整参数,生成PSO-DBN深度学习网络模型,对比改进型BP神经网络识别方法,避免了随机初始化导致的局部最优解问题,粘接质量识别的准确度有所提高。

Aiming at the problem of identifying the bonding quality of thin plate composite materials,this paper proposes an improved deep confidence network(DBN) deep learning model based on particle swarm optimization(PSO). PSO is used to optimize the weight of standard DBN network to solve the problem of classification overfitting in the identification process of thin plate composite materials, and the identification process of the bonding quality of thin plate composite materials is completed by using this model. It discusses the design steps of the algorithm, debugs and trains the model for many times,determines the parameters of the network model, and completes the identification of the test data. The simulation results show that the pso-dbn deep learning network model is generated after many training tests and adjusting parameters. Compared with the improved BP neural network identification method,the local optimal solution problem caused by random initialization is avoided, and the accuracy of adhesive quality identification is improved.