基于集合极限学习机的模拟电路问题处理

Processing of Analog Circuit Problem Based on Ensemble Limit Learning Machine

针对极限学习机由于初始参数随机生成引起的诊断结果不稳定问题,提出一种集合极限学习机的模拟电路故障诊断方法.首先利用数个单独极限学习机分别建立诊断模型,得到对应的诊断结果;然后将各极限学习机的输出值进行等权平均处理,得到集成后的输出结果,依此确定故障类别;最后,将方法应用于单管共射极放大电路故障诊断.结果表明:集成极限学习机相比传统神经网络和单个极限学习机,诊断精度最高,达到了99.66%,同时结果稳定性得到了有效提高,说明了方法用于故障诊断的有效性.

Aiming at the problem of unstable diagonizing results caused by randomly generated initial parameters in extreme learning machine（ ELM）,an analog circuit diagnosis method based on extreme learning machine is proposed.A couple of single ELMs are used to set up the diagnosis model to get diagnosis results.Then,output values of every single ELM are processed with equal weight to confirm fault categories.Lastly,the proposed method is used for fault diagnosis of single transistor common emitter amplifier.Results show that diagnosis accuracy of ELM can reach 99.66%,which is higher than traditional neural networks and single ELM.