Fault detection and diagnosis (FDD) system for end of production line testing of alternators

产品质量和可靠性仍然是汽车制造商的首要任务之一，尽管面临竞争压力，需要更高的制造效率和更低的生产成本，但必须保持这一点。随着原始设备制造商(OEM)转向更精简的结构，对其设备供应商确保质量和可靠性的要求越来越高。通常，汽车供应商的内部资源比OEM少，执行所需的质量和可靠性分析的能力和专业知识较少。本提案是D&V Electronics与D&V Electronics关于开发交流发电机生产线末端测试的故障检测和诊断(FDD)能力的项目的继续。到目前为止，这种合作已经导致了一种被称为工业扩展多尺度原理元件分析(IEMSPCA)的算法的开发，该算法以前已应用于汽车起动器。在IEMSPCA中实现了振动和声音的测量。它在不到15秒的时间内进行故障诊断分析，在26377个起动机测试案例中应用于已知故障时，检测成功率达到96%以上。在这个项目中，IEMSPCA将得到进一步开发，能够检测、诊断、表征和适应发电机线端测试中发生的已知和先前已知的故障情况。将应用深度学习等智能策略，并将使用更大的信号阵列，包括振动、声音、电压和电流测量。将应用基于模型的策略来提高对选定范围的故障条件的诊断能力。研究成果将在ad&V电子测试单元平台上实现。

Product quality and reliability continues to be one of the top priorities for auto manufacturers, and these mustbe maintained despite competitive pressures and the need for higher manufacturing efficiency and lowerproduction costs. As the Original Equipment Manufacturers (OEMs) move to leaner structures, the demandsplaced on their equipment suppliers to assure quality and reliability is increasing. Typically the automotivesuppliers have fewer internal resources than OEMs with less capability and expertise to perform the requiredquality and reliability analysis.This proposal is a continuation of a project with D&V Electronics on developing Fault Detection andDiagnosis (FDD) capabilities for end of production line testing of alternators. This cooperation has so farresulted in the development of an algorithm referred to as the Industrial Extended Multi-Scale PrincipleComponents Analysis (IEMSPCA) that has previously been applied to automotive starters. Both vibration andsound measurement were implemented in IEMSPCA. It performed FDD analysis in less than 15 seconds andhas achieved more than 96% detection success rate when applied to known faults on 26377 starter test cases.In this project the IEMSPCA will be further developed to be able to detect, diagnose, characterize and adapt tounknown as well as previously known fault conditions occurring during end-of-line testing of alternators.Intelligent strategies such as deep learning will be applied and a larger array of signals including vibration,sound, voltage, and current measurements will be used. Model-based strategies will be applied for improvingdiagnosis capabilities for a select range of fault conditions. The research outcomes will be implemented on aD&V Electronics test cell platform.