论文部分内容阅读
激光焊接过程复杂,影响因素众多,许多参数难以量化。基于归一化的径向基函数神经网络,采用非参数统计方法,建立了激光焊接熔池在时间和空间上的光强分布模型。该神经网络采用高斯函数作为径向基函数。提出了定量评价该模型预测光强分布质量的方法,并根据该评价方法,对影响光强分布模型的重要参数进行优化选择。根据优化选择结果,对两幅光强分布图形进行预测。通过预测图像与实测图像的对比证明,该神经网络可有效预测激光焊接熔池的光强分布。
Laser welding process is complex, many influencing factors, many parameters difficult to quantify. Based on the normalized radial basis function neural network, a non-parametric statistical method was used to establish the light intensity distribution model of the laser welding pool in time and space. The neural network uses Gaussian function as a radial basis function. A method for quantitatively evaluating the quality of the light intensity distribution of the model is proposed. According to the evaluation method, the important parameters affecting the light intensity distribution model are optimized and selected. According to the results of optimization, two light intensity distributions are predicted. The comparison between the predicted image and the measured image proves that this neural network can effectively predict the light intensity distribution of the laser welding pool.