Bradley Walsh, & Dr. Vida Vakilian
Significance of Study
This research addresses a critical challenge in the field of automation: the inevitable wear and tear of machinery which necessitates costly and time-consuming maintenance and repairs. The significance of this project lies in its potential to revolutionize the approach to machine maintenance, shifting from reactive to proactive measures that can minimize both the financial burden and downtime associated with traditional preventative maintenance and improve the reliability and efficiency of automated systems.
Subject Background
Machinery in automated systems is prone to degradation over time, leading to operational inefficiencies and potential failures. Traditional maintenance strategies are often reactive and can be prohibitively expensive and disruptive. The theoretical foundation of this research is built upon the premise that machine learning models, particularly those capable of detecting anomalies, can play a pivotal role in identifying and predicting potential issues before they result in significant downtime or costly repairs. Prior explorations into models such as long short-term memory (LSTM) recurrent neural networks and k-nearest neighbors (KNN) classifiers have provided valuable insights, yet they fall short in accurately and reliably identifying anomalies compared to the isolation forest model.
Methodology
The research focuses on the development and testing of an isolation forest machine learning model. This model will analyze data points from sensors embedded in automation machinery to identify potential anomalies indicative of wear or failure. Comparative analysis will be conducted with other machine learning models, such as LSTM recurrent neural networks and KNN classifiers, to evaluate the efficacy and reliability of the isolation forest model in real-world scenarios. The study will also investigate the challenges of integrating machine learning models into existing automation systems, particularly the limitations of programmable logic controllers and the capacity of attached computers to process and act on the data in real time.
Preliminary Findings
While the research is ongoing, preliminary findings suggest that the isolation forest model shows promise in accurately detecting anomalies that precede mechanical failures. Expected findings include a detailed analysis of the model’s performance compared to other machine learning approaches, highlighting its superiority in specific contexts of anomaly detection.