基于改进相关向量机的模拟电路故障预测

ANALOG CIRCUIT FAULT PROGNOSTIC APPROACH BASED ON IMPROVED RELEVANCE VECTOR MACHINE

针对模拟电路实际使用过程中的可靠性问题,提出一种基于改进相关向量机的模拟电路故障预测方法。在提取被测电路的故障特征参数之后,将传统相关向量机中单个核函数改进为多个不同核函数组合形成混合核函数,提高算法的泛化性能。利用量子方法改进人工蜂群算法提升其优化效果以对各个核函数的权重因子寻优,从而提高算法的预测精度。将离散灰色模型与多核相关向量机相结合,提升算法的长期趋势预测性能。仿真结果表明,该方法的绝对误差、相对误差和测试误差均小于传统的相关向量机预测方法。

Aimed at the reliability problems in the actual use of analog circuits,an analog circuit fault prognostic approach based on improved relevance vector machine is proposed.After extracting the fault characteristic parameters of the circuit under test,a single kernel function in traditional relevance vector machine was improved into a combination of multiple different kernel functions to form a mixed kernel function,which enhanced the generalization performance of the algorithm.The quantum method was used to improve the artificial bee colony algorithm to improve its optimization effect and optimize the weighting factors of each kernel function,so as to enhance the forecasting accuracy of the algorithm.The discrete gray model and multi-core correlation vector machine were combined to enhance the long-term trend prediction performance of the algorithm.The simulation results show that the absolute error,relative error and measure error of this method are less than the traditional relevance vector machine prognostic approach.