基于改进神经网络算法的医疗锂电池PHM系统设计

Electromedical Lithium Battery PHM System Based on Improved Neural Network Algorithm

摘要:

针对医疗电子设备锂电池不确定性发生故障耽误病人救治的问题,提出了一套医疗电子设备锂电池故障预测与健康管理系统(Prognostics and Health Management-PHM)。搭建了一套医疗电子设备锂电池数据测试与退化状态模拟的实验平台。为了反映医疗电子设备锂电池健康状态,将锂电池四个健康因子作为医疗电子设备锂电池退化状态的特征进行提取,并通过非线性自回归(Nonlinear Autogressive with Exogenous Inputs-NARX)神经网络,对四个健康因子的数据进行训练,训练后用于容量估计,得出等间隔放电时间序列能够较好地表征锂电池健康状态。为了提高基本粒子滤波算法(Particle Filter-PF)的精度从而更精确地预测锂电池剩余寿命(Remaing Useful Life-RUL),通过人工免疫粒子滤波算法(Artificial Immune Particle Filter-AIPF)与经验模型对锂电池进行剩余寿命预测,并将PF预测的结果与AIPF预测的结果进行对比,发现AIPF预测更加准确,说明AIPF有效抑制了PF重采样过程中粒子退化问题,验证了医疗电子设备锂电池故障预测与健康管理系统的可行性与可实施性。

Abstract:

In order to solve the problem of failure of patients with failures caused by the uncertainty of lithium-ion batteries in medical electronic equipments, a set of prognostics and health management (PHM) systems for lithium-ion batteries in medical electronic equipment was proposed. An experiment platform for data testing and degradation status simulation of lithium batteries for medical electronic equipment was built. In order to reflect the health status of lithium-ion batteries for medical electronic devices, the four health factors of lithium batteries are extracted as characteristics of the degradation status of lithium-ion batteries for medical electronic devices, and they are passed through a nonlinear auto-regressive with exogenous inputs (NARX) neural network. The data of the health factors were trained and used for capacity estimation after training, and the equal interval discharge time series could be used to better characterize the lithium battery health status. In order to improve the precision of the Particle Filter-PF and more accurately predict the Reamaling Useful Life-RUL, the Artificial Immune Particle Filter (AIPF) and the Empirical Model for Lithium The battery performs the remaining life prediction, and compares the PF prediction result with the AIPF prediction result, and finds that the AIFF prediction is more accurate, indicating that AIFF effectively inhibits the particle degradation problem in the PF re-sampling process, and verifies the failure prediction of the lithium ion battery for medical electronic equipment. Health management system feasibility and enforceability.