Explaining Anomalies in Industrial Multivariate Time-series Data with the help of eXplainable AI

The large amount of data generated by industrial plants provides an excellent opportunity to use Machine Learning (ML) for a better understanding of plant behaviour. Thus, supporting plants operators in running their plants efficiently. An example of an anomaly detection method is to help plant operators detect if their plant is still running normally or if curative actions are needed. Users of ML-based solutions often complain of a challenge called lack of interpretability, i.e., the degree to which one can understand the model's outcome. To address the need for better interpretability of ML models, the re-search field eXplainable Artificial Intelligence (XAI) has recently received increasing attention in the industrial domain. This paper performs a survey and investigates different XAI techniques that can be applied for detecting anomalies in industrial plant assets. The paper focuses on multivariate time-series data because a) it is the most predominant type of data in industrial systems and b) all the available XAI techniques have applications for various data types such as images, tabular, or textual data. However, these techniques are generally not well suited to explain anomalies in multivariate time-series. To solve this problem, we build an anomaly detection method for the multivariate time-series data generated by the industrial simulators using auto-encoders. Based on feature attribution, examples, and trees, seven different XAI techniques are ideated, developed, and discussed. Out of the seven XAI techniques, a SHAP based explainer, called DTFS, correctly identified the root cause of the anomaly with an accuracy of 86% and took 1.53 seconds to explain our benchmark system.

I. Introduction

Modern industrial process plants such as chemical plants or oil-refineries collect a vast amount of data [1], which provides a great opportunity for Artificial Intelligence (AI) and Machine Learning (ML) [2]. Various possible ML use cases may apply, however, one of the most important use cases is anomaly detection [3]–[6]. Due to advances in automation technology, the trend today for operators is to observe the automation, ensure safety, and “normal” running of the process plant, instead of manually operating the process. Monitoring an industrial process plant is challenging and depends on the operator's experience. Handling a critical situation requires quick analysis and reaction that can be challenging even for experienced operators. Suppose the production is no longer within normal bounds. In that case, the operator needs to analyse the problem and take appropriate actions. Anomaly detection provides this kind of information to the operator, however, for two reasons, anomaly detection is not an entirely satisfying solution for industrial practice. First, the lack of variability in the process manufacturing system provides only training data sets within a narrow operational range, i.e., most situations are labelled as acceptable, and a few rare situations will be labelled anomalous. Second, the anomaly detection methods usually only point the operator to the anomaly. However, due to the complexity of plant equipment, there is often the need for a better explanation to help the operator in the situation analysis [7].