基于PSO-BP分区处理的电路元件过热检测方法

Based on PSO and BP torn parallel processing of analog circuit component element overheat detection method

研究模拟电路元件过热检测准确率优化问题,传统的方法计算量大、实时性差、元件过热检测的准确率不高,提出基于优化神经元件网络分区并行处理的模拟电路元件过热检测方法,由于模拟电路元件的热量输入输出量是连续变量,元件之间的离散性使得电路过热等效于某个元件的过热;利用此特点,将模拟电路元件按照交叉原则分区成若干元件小元件网络,将元件小元件网络的特征信息并行输入神经元件网络过热检测模型,使用粒子群优化的方法对神经元件网络进行参数优化,防治检测陷入局部最优化,以提高电路元件过热检测的准确率;在不同的元件模拟电路中进行测试,测试结果证明,该方法对模拟电路元件的过热检测的准确率高达92%,具有很强的实用性。

The analog circuit components overheat detection accuracy optimization problem. Traditional method of large amount of calcu- lation and poor real--time performance, components overheat detection accuracy is not high. Is put forward based on the optimization of neu- ral network partition parallel processing components of analog circuit components overheat detection method, due to the amount of heat input and output analog circuit component is a continuous variable, element discreteness between circuits has overheating is equivalent to a certain element of overheating. To the feature of the analog circuit element partition into several components small components in accordance with the principle of cross network, the components, the feature information of the small element network paraIlel input neural element network overheat detection model, using particle swarm optimization method of neural network of units and parameter optimization, control testing trapped in local optimum, in order to improve the circuit components overheat detection accuracy. In different components in analog circuit test, the test results show that the method of analog circuit components overheat detection accuracy is as high as 92~, have very strong prac- ticability.