Bayesian Sensor Estimation for Machine Condition Monitoring

RelSen: An Optimization-based Framework for Simultaneously Sensor Reliability Monitoring and Data Cleaning

“…where an Expectation-Maximization (EM) algorithm [3] is used to estimate z t and Θ t simultaneously. More details of the BayesGMM method can be found in [32]. Note that the BayesGMM method does not explicitly utilize sensor redundancy for process estimation, as a result, let z t s denote the estimated process state for sensor s, we calculate z t p = s ∈S p z t s .…”

“…However, as the ground truth of the measured signals is generally unknown, sensor data cleaning methods have to be applied beforehand. Among the common approaches for sensor data cleaning are state estimation methods [2,6,7,25], parameter estimation methods such as Bayesian estimation algorithms [32,33] and truth estimation algorithms [26], and multi-sensor fusion techniques [5,29]. To formulate the whole workflow, the authors of [27] proposed a sensor accuracy estimation framework which consists of four layers: pre-processing, state estimation, accuracy estimation and accuracy indexing.…”

“…These techniques can be broadly categorized into three classes: 1) state estimation methods which estimate the states of monitored processes by utilizing the prior knowledge about the process state transition dynamics and the distribution of measurement noise. The family of Bayesian filters [1,6,7] are within this class; 2) parameter estimation methods which estimate the parameters of the sensor measurement errors that best describe the observed sensor measurements via the learned sensor models, examples see [26,28,32]; 3) multi-sensor fusion techniques which combine measurements from redundant sensors to achieve improved measuring accuracy than that could be achieved by the use of a single sensor alone. Typical fusion methods combine measurements from multiple sensors using mean, median or weighted average statistics based on known noise variance or covariance of sensors [24,31].…”