AI-Driven Maintenance Algorithms For Intelligent Network Systems: Leveraging Neural Networks To Predict And Optimize Performance In Dynamic Environments

Abstract

The increasing levels of intelligent technology to be incorporated into both physical systems and the systems’ internal ‘intelligent’ functionalities represent a major frontier in future networked intelligence science and technology innovations. Technical performance and reliable operation in ever more complex environments will require correspondingly higher levels of intelligent inputs and feedback. This paper explores the use of evolving neural networks, combined with more traditional techniques, to make physical system maintenance a vital feedback mechanism for the sustainable and reliable delivery essential for new intelligent network systems. New AI/ML and neural network maintenance capabilities can work on micro- and macro-scales across an intelligent network environment to predict and monitor potential equipment failures. They also enable efficient maintenance scheduling for local equipment supplies or for supply lines of equipment availability that must be carried out while protecting an external service from network interruptions.
Our model is based on the fusion of multi-resolution high-frequency data, or dynamic high-resolution data analysis of data collected from the environment depicting the impact on cyber-physical systems caused by system operation. Changes and anomalies in this data are interpreted and enabled with the use of machine learning techniques for prediction, anomaly detection, and model quality assessment. The goal is to predict different types of hazards related to component deterioration using these proposed depth features. We also explore both time-honored and the leading edge of anomaly and deterioration predictor pattern identification in labels that are transformative. Entropy values are determined for anomaly detection, and our proposed model's effectiveness is validated through real-world scenario simulation including root cause analysis. We present visualization tools as decision support for field technicians to understand and monitor the results. Potential future work for expanding the types of data is also discussed in the conclusion.