Time-Shifted Maps for Industrial Data Analysis: Monitoring Production Processes and Predicting Undesirable Situations

The rapid advancement of computing power, combined with the ability to collect vast amounts of data, has unlocked new possibilities for industrial applications. While traditional time-domain industrial signals generally do not allow for direct stability assessment or the detection of abnormal situations, alternative representations can reveal hidden patterns. This paper introduces time-shifted maps (TSMs) as a novel method for analyzing industrial data-an approach that is not yet widely adopted in the field. Unlike contemporary machine learning techniques, TSM relies on a simple and interpretable algorithm designed to process data from standard industrial automation systems. By creating clear, visual representations, TSM facilitates the monitoring and control of production process. Specifically, TSMs are constructed from time series data collected by an acceleration sensor mounted on a robot base. To evaluate the effectiveness of TSM, its results are compared with those obtained using classical signal processing methods, such as the fast Fourier transform (FFT) and wavelet transform. Additionally, TSMs are classified using computed correlation dimensions and entropy measures. To further validate the method, we numerically simulate three distinct anomalous scenarios and present their corresponding TSM-based graphical representations.