Fault Detection and Diagnostics Using the Mahalanobis-Taguchi System

Abstract

SUMMARY & CONCLUSIONSFault detection and diagnostics (FDD) refers to identifying the faulty operations of a machine and then determining the cause or causes of the fault. FDD is an important use case of Industrial Internet of Things (IIOT) technology. This paper introduces the Mahalanobis-Taguchi System (MTS) method, which is an important technique used in FDD. The MTS method was proposed by Genichi Taguchi for diagnosing multivariate data. It has been successfully used for FDD and quality control in a variety of industries, including electrical power, chemical, and aerospace. The MTS method involves the following three steps.Step 1: First, the multivariate sensor data from a machine’s normal operations (or fault-free operations) are collected. The sensor variables are standardized, and the Mahalanobis distance (MD) of the observations from the center is calculated. These MD values are used to define what is referred to as the Mahalanobis space of the normal observations. The MTS method uses the Mahalanobis space as a scale of reference for identifying outliers. This step can be considered the model-training step.Step 2: Second, new observations are scored by calculating their MD values. During scoring, the observations are first scaled using the mean and standard deviation of sensor variables by using the machine’s normal operations data, which were collected in step 1. The MD computation in this step uses the correlation matrix of the normal operations data. An observation is considered to be an outlier, or the presence of a fault is indicated, if that observation’s MD value is higher than the typical MD value of observations in normal or fault-free operations.Step 3: Third is the model interpretability step, in which fault diagnosis is performed by identifying important variables. For such a determination, a design-of-experiments approach is adopted using orthogonal arrays and the signal-to-noise (S/N) ratio.Because the MTS method is an interpretable outlier detection technique, it is very attractive for FDD. A machine operator can use the output of the MTS method and take timely corrective actions. This paper introduces the MTS method for FDD and evaluates this method by using several datasets. We include a detailed discussion of the advantages of the method, as well as its limitations. Then we introduce a new kernel-based MTS method that can be useful when data from normal operations of the machine are multimodal and/or non-Gaussian.