Having an integrated health management and diagnostic strategy becomes an important part of a system’s operational life-cycle. This is because it can be used to detect anomalies, analyze failures and predict the future state of the APC valve based on currently collected information. By utilizing condition data and on-test system feedback, data models can be trained using machine learning and statistical concepts. Once trained, the logic for data processing can be embedded on our-board controllers to enable real-time health assessment and analysis.
This concept continues to face several challenges, but some pieces of the puzzle are coming together. First, to engage in this process one must first have an embedded platform with a high level of processing power. In our case the A15 processor fitted with multicore deep learning accelerators (DLA) co-processors. Second, while the inference can be done at the edge, the training phase must be done offline in powerful workstation through extensive data collection programs in which we analyze, transient Vibration signatures, valve movements profiles at the micro level, as well as motor torque profiles. We currently have 3 test machines attempting to collect a data set of what requires classification or detection to develop a trained network.
We are striving to continue putting the pieces of the puzzle together in order to help this state-of-the-art technology become efficient in real-time use cases particularly for active performance optimization of our APC valves. Hereby to the left is our first Machine Learning Capable control unit. This unit is integrated to APC valve hardware of different configuration and style.