Development of IoT-enabled Predictive Maintenance System for Industrial Machinery

Thesis Abstract

In the contemporary industrial landscape, unplanned machinery downtime and maintenance costs significantly impede operational efficiency and competitiveness, underscoring the necessity for proactive maintenance strategies. This study aims to develop an Internet of Things (IoT)-enabled predictive maintenance system that enhances the accuracy and timeliness of fault detection in industrial machinery. The specific objectives include (1) designing an IoT-based sensor network for real-time data acquisition; (2) developing a predictive model utilizing machine learning algorithms to forecast machinery failures; and (3) evaluating the system's effectiveness in reducing downtime and maintenance costs within an industrial setting. The research adopts a mixed-methods approach, combining quantitative data analysis with qualitative system evaluation. The study population comprises five key manufacturing firms operating in the automotive parts sector, with a purposive sample of 50 industrial machines selected based on criticality and operational history. Data collection involves deploying IoT sensors measuring vibration, temperature, and operational humidity over a six-month period, complemented by maintenance logs and failure reports. The IoT data are preprocessed and analyzed using regression analysis and support vector machine (SVM) classifiers to develop a predictive maintenance model, while qualitative feedback from maintenance personnel is analyzed thematically to assess system usability and integration challenges. This research anticipates findings that demonstrate a significant correlation between sensor data patterns and impending machinery failures, with the predictive model achieving an accuracy rate of approximately 85% in fault detection. The implementation of the IoT-enabled system is expected to facilitate maintenance scheduling that reduces unplanned downtimes by up to 30% and maintenance costs by 20%, thereby increasing overall operational efficiency. The study contributes to existing knowledge by integrating sensor-driven data analytics with scalable IoT architectures specifically tailored for industrial maintenance, extending the theoretical framework of the Technology Acceptance Model (TAM) and Reliability Theory to the context of predictive maintenance systems. Additionally, it offers empirical evidence on the efficacy of machine learning algorithms in operational predictive analytics, addressing gaps identified in prior studies that primarily emphasize isolated sensor technologies or limited case analyses. The main conclusion affirms that IoT-enabled predictive maintenance systems substantially improve machinery reliability and operational productivity in manufacturing environments. Recommendations include strategic integration of IoT infrastructure with existing enterprise resource planning (ERP) systems, staff training on sensor data interpretation, and the development of standardized protocols for sensor deployment and data security. The study further suggests avenues for future research, such as exploring advanced deep learning models for failure prediction and expanding the scope to include multi-site industrial deployments. Overall, the research establishes a robust framework for leveraging IoT technologies in industrial maintenance, emphasizing that the fusion of sensor data analytics and predictive modeling can transform reactive maintenance paradigms into proactive, data-driven decision-making processes that significantly enhance industrial competitiveness.