Digital Twin Technology Approach Based on the Hierarchical IDP Tensor Decomposition

Digital Twin (DT) technology is based on the idea for the creation of a digital model which accurately represents a physical object, or a process. At first, the studied object is equipped with sensors, specially selected to account for and follow the changes of the most important functions, which provide information of various kinds. This information could be a big amount of data, related to different aspects of investigated process/object performance, which is then transferred to a processing system. The so collected data is suitably transformed, and then—used to create the digital model (i.e., the Digital Twin), which is after that used for comparison, evaluation and control, depending on the application. This virtual representation of an object or a system could be updated by using various additional information tools, such as real-time data, and together with this, simulation, machine learning and reasoning which help the decision-making in the system performance. The Hierarchical Inverse Difference Pyramid (IDP) Tensor Decomposition suits extremely well the DT technology. For this, the virtual model is represented as a multilevel tensor. Then, the comparison and the analysis of the investigated process are executed following the IDP decomposition levels, which offer increasing correspondence and similarity together with the hierarchical level number increase. The proposed structure is flexible and permits various approaches: (1) In case that a process is investigated, its execution could be followed in real-time. In each decomposition level are preset the normal parameters and threshold values, which permit the detection of possible deviations, on the basis of which, the process to be corrected. Additional advantage is the possibility to detect the most significant trends in the investigated system performance. (2). In case that an object is searched in a large image database, similar approach is used, based on the similarity evaluation in the consecutive tensor decomposition levels. The corresponding block diagrams are given in the paper.