Experimental and Simulation analysis of Monitoring a Industrial process by Adaptive transfer Learning
Authors:
- 27 October 2023
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 1, 1-7
Abstract
A wide range of systems can be monitored using data-driven processes, and reliance on response mechanisms and first principles has decreased for complex manufacturing schemes. Most data-driven approaches to process monitoring, however, use the assumption that data from previous training sessions and data from online testing sessions have the same distribution. The hostile surroundings of industrial systems constantly affects the gathered information from production operations, including the ever-changing working scenario, variation in the original materials, and manufacturing indices. Due to these reasons, there is frequently a discrepancy between the live monitoring data distribution and the historical learning data distribution, which leads to a faulty model for the process monitoring job. This means it takes more work to get reliable results in process monitoring by using a model gleaned from historical data. This research proposes the use of a Resilient Transfer Dictionary Learning (RTDL) technique in industrial process monitoring. This technique addresses the discrepancy between historical training data and online testing information caused by shifting operational conditions. Domain-adaptive transmission and Representative learning come together to form the RTDL. In the context of the transfer learning issue, the suggested approach considers the source domain, the historical training data and the target domain, the online testing data. There are two types of regularization for dictionary learning: linear discrepancy and maximum mean discriminant analysis Regularisation helps to reduce the distributional gap between source and target domains. Under a realistic and dynamic operational environment, this method may be used to develop a robust vocabulary even if the source domain and goal area features are different. Process monitoring and mode categorization are two areas that might benefit greatly from such a lexicon. The effectiveness and superiority of the suggested technique are tested by extensive tests using a numerical simulation and two industrial systems.Keywords
This publication has 20 references indexed in Scilit:
- Surviving Disaster Events Via Dynamic in-Network Processing Assisted by Network Digital Twins2023
- Multiple structured latent double dictionary pair learning for cross-domain industrial process monitoringInformation Sciences, 2023
- Unsupervised Domain Adaptation via Subspace Interpolating Deep Dictionary Learning: A Case Study in Machine Inspection2023
- Dictionary learning-based damage detection under varying environmental conditions using only vibration responses of numerical model and real intact State: Verification on an experimental offshore jacket modelMechanical Systems and Signal Processing, 2022
- Women's Safety with a Smart Foot Device2021
- A hybrid approach estimates the real-time health state of a bearing by accelerated degradation tests, Machine learning2021
- Cross-domain EEG signal classification via geometric preserving transfer discriminative dictionary learningMultimedia Tools and Applications, 2021
- 4D seismic history matching: Assessing the use of a dictionary learning based sparse representation methodJournal of Petroleum Science and Engineering, 2020
- Nonlinear process monitoring using kernel dictionary learning with application to aluminum electrolysis processControl Engineering Practice, 2019
- A Survey of Dictionary Learning Algorithms for Face RecognitionIEEE Access, 2017