Advancing Machinery Predictive and Prescriptive Maintenance Systems Using Artificial Intelligence A Study and Evaluation

Abstract

Maximizing machine uptime is crucial, especially when repair durations impact production newlinecontinuity. Manufacturing and assembling machine components are essential newlinefor operational efficiency, with ongoing maintenance playing a vital role in sustaining newlineequipment health. Predictive maintenance strategies, which leverage real-time operating newlineconditions and tool data, are widely adopted to support this. In industrial settings, newlineunexpected machine downtime often leads to production bottlenecks, safety risks, and newlineworkflow disruptions. Over time, these issues contribute to reduced output, increased newlineoperational costs, and significant revenue losses. Therefore, minimizing downtime and newlineswiftly restoring failed equipment to production is of paramount importance. newlineOne significant application of Machine Learning and Deep Learning in the industrial newlinesector is predictive maintenance. This involves using a machine learning model newlineto anticipate equipment failure based on operating parameters. Techniques like analyzing newlinevibration signals to identify anomalies, such as surface defects in ball screws, newlineare commonly adopted. However, this domain is complex due to the wide variation newlinein machinery design and operating environments, posing a challenge to developing a newlinestandardized predictive model. Moreover, predictive maintenance alone does not guarantee newlinequicker restoration of equipment to operational status, as human intervention time newlineremains a critical factor. Current methodologies employ deep learning techniques integrated newlinewith ensemble frameworks to detect faults in ball screws. Despite their effectiveness, newlinethese approaches have high computational time, making the process slow newlineand inefficient. Additionally, some studies focus primarily on fault detection without newlineaddressing defect severity classification, which is crucial for prioritizing maintenance newlineactions. Techniques like Fuzzy Analytic Hierarchy Process (AHP) combined newlinewith the technique for Order Preference by Similarity to Ideal Solution (TOPSIS) have newlinebeen applied to identify optimized