<b>Prognostics and Fault Diagnosis </b><b>for </b><b>Manufacturing Equipment </b><b>using Machine Learning </b>

<p dir="ltr">With the advancement of computing and sensor technologies, smart manufacturing has gained significant attention in the realm of Industry 4.0. Machine learning (ML) and Artificial Intelligence (AI) has accelerated the transition of traditional manufacturing technologies into “smart” technologies. Their ability to handle tremendous amounts of data, make real-time decisions, and address more complex manufacturing problems has facilitated various fields in both academic and industrial applications.</p><p dir="ltr">Maintenance plays a critical role in manufacturing systems, ensuring operational efficiency and reliability of the system. However, traditional maintenance activities involved significant human resources, resulting in human errors and inefficiency. Additionally, it is a costly and time-consuming process to sufficiently train human maintenance engineers. Hence, data-driven technologies such as ML and AI have been adopted as more advanced alternatives for manufacturing technologies.</p><p dir="ltr">Predictive maintenance has been extensively studied as advanced data-driven technologies enabling smart manufacturing. In factories, maintenance practices are evolving from mere equipment “management” to equipment “wellness,” creating an integrated and smart manufacturing system that responds in real-time to changing conditions. Equipment wellness involves being aware of the equipment's health condition and making-decisions to secure operational and system-level requirements. This requires analyzing large amounts of machine condition data obtained from sensors to diagnose current health conditions and predict future behavior, such as remaining useful life. If a fault is detected, identifying its root cause is essential for extending equipment life and preventing recurrence.</p><p dir="ltr">However, developing a model that captures the relationship between multi-sensor signals and mechanical failures is challenging due to the dynamic manufacturing environment and the complexity of mechanical systems. Another key challenge is obtaining usable machine condition data to validate the method.</p><p dir="ltr">The proposed work aims to develop a systematic tool for maintenance in manufacturing plants using emerging technologies such as AI, Smart Sensors, and IoT. This method will aid decision-making by quickly detecting worn components and estimating remaining useful life. To diagnose and prognose equipment health, several data-driven models will be developed and validated through experiments across various manufacturing scenarios, such as cutting tools, gears, and bearings. Signal processing will preprocess raw signals using domain knowledge, extracting and selecting useful features to enhance computational efficiency in model training. Customized deep learning architectures will then be designed to effectively and efficiently learn the relationship between processed signals and model outputs, such as health indicators. Ultimately, this research aims to prevent catastrophic mechanical failures, improve product quality, and extend equipment service life.</p>