Intelligent diagnostics and prognostics for machinery systems

生产质量、运营安全和经济状况直接影响到加拿大组织在全球市场上的竞争力。工业中迫切需要一个可靠的在线监测系统，以提供潜在损害的早期预警，从而防止机械性能下降、故障和灾难性故障。经典监测系统的主要缺点是缺乏可靠性和稳健性，特别是在运行条件时变的情况下。这项研究的目标是开发新一代智能诊断和预测(IDP)系统，用于在线、更可靠地监测机械的健康状况。当这样的IDP系统被实施时，它将能够识别机器的健康状况。当潜在问题出现时，IDP系统可以精确定位故障部件，估计故障传播趋势，并预测损坏机组的剩余使用寿命。这项研究的第一个重点是开发一种新的技术来提取特定的旋转波形来检测旋转机械的轴承故障。文献中提出了许多轴承故障检测技术，但每种技术都有各自的优点和局限性。第二个目标是全面调查所有经典和建议的信号处理技术在不同条件下用于轴承和电机故障检测的稳健性。开发了一种新的转换技术来将基于模型的范例映射到神经模糊原型，以努力进一步提高所得到的方案的性能。提出了一种IDP智能工具，以有效地集成来自分类器和预测器的处理信息，以提供对机械健康状况的更积极的评估。将制定在线/离线培训的新标准和战略，以在实时操作期间提取新知识，以提高监测系统的适应能力和稳健性，并适应不同的机械条件。

Production quality, operational safety, and economics have a direct impact on the competitiveness in the global market of organizations in Canada. A reliable online monitoring system is crucially needed in industries to provide an early warning of potential damage that will subsequently prevent machinery performance degradation, malfunction, and catastrophic failure. The principal disadvantages of classical monitoring systems are the lack of reliability and robustness, especially for time-varying operating conditions. The goal of this research is to develop a new generation of intelligent diagnostic and prognostic (IDP) systems for online, more reliable monitoring of the health conditions of machinery. When such an IDP system is implemented, it will be able to recognize the health conditions of machinery. When a potential problem arises, the IDP system can pinpoint the faulty components, estimate the fault propagation trend, and forecast the remaining useful life of the damaged unit. The first focus of this research is to develop a new technique to extract specific rotation waveforms to detect bearing faults in rotary machinery. Many techniques have been proposed in the literature for bearing fault detection; however, each has its own merits and limitations. The secondary objective is to comprehensively investigate the robustness of all classical and proposed signal processing techniques for bearing and motor fault detection corresponding to different conditions. A new transformation technique has been developed to map a model-based paradigm to a neural fuzzy prototype in an effort to further improve the performance of the resulting scheme. An IDP intelligent tool is proposed to effectively integrate the processing information from both the classifier and the predictor to provide a more positive assessment of the health conditions of machinery. New criteria and strategies for online/offline training will be developed to extract new knowledge during real-time operations that will increase adaptive capabilities and robustness of the monitoring systems and accommodate different machinery conditions.