Control signal based fault detection in streaming employing Principal Component Analysis and soft sensor in linear actuators

The complexity of industrial systems has increased significantly, with many interconnected components, multiple control mechanisms, and a wide range of process variables and operational data. It is crucial to keep track of how these systems are functioning, as any deviation from normal operation due to a malfunction can result in significant financial losses and safety risks. Linear actuators are an essential component in many industrial and technological applications. Used for a wide range of tasks, such as moving, positioning, and controlling objects, they are a key component in many automated systems. This study presents a proposal for fault detection in streaming in linear actuators using Principal Component Analysis (PCA), Hotteling T 2 ( T 2 ), with a Bayesian soft sensor setup. Two different setups were tested: using a fixed threshold, and with an adaptive threshold. Results show that the two strategies were able to identify the three categories of faults, Lack of lubrication, Backlash, and Spalling in some degradation stage or load.