The Challenge: Live Fault Detection Without Disturbance
The primary objective of this research was to develop a live fault detection system within industrialized winding processes for small motors, utilizing an AI approach. A critical aspect of this endeavor was to ensure that the AI implementation did not disturb the ongoing industrial process. This delicate balance was achieved by designing a system that could record and store process data, and then apply AI analysis without interfering with the machinery’s operation.
Innovative Approach: Raspberry Pi and Deep Learning
The heart of this innovation lies in its use of a Raspberry Pi for computational resources, ensuring live anomaly detection. This approach not only proved to be efficient but also cost-effective. We adopted a two-pronged strategy in developing the AI models:
-
Binary Classification: This model was tasked with determining whether an error occurred. Although it faced challenges with certain error patterns, adjustments in the approach boosted its accuracy to nearly 90%.
-
Categorical Classification: More advanced than its counterpart, this model could predict the type of error with an impressive accuracy of 92.5%. It utilized a combination of convolutional neural networks and long short-term memory networks, striking the right balance between complexity and performance.
Practical Implementation and Results
The implementation process was meticulously planned and executed. Initially, the team manipulated the machine to generate different types of errors, creating a rich dataset for training the AI. This dataset included four distinct error types: incorrect wire usage, closed hook errors, wrong loading heights, and angle positioning errors.
Upon successful training and implementation, the AI demonstrated its ability to work in real-time, matching the production line’s speed without any lag. This achievement is not just a testament to the Raspberry Pi’s capability but also to the efficient AI model developed by the researchers.
Conclusion: A Milestone in Industrial AI Applications
This research marks a significant milestone in the application of AI in industrial processes. It showcases how AI can be seamlessly integrated into existing manufacturing setups, enhancing efficiency and precision without incurring high costs or operational disturbances. The success of this project paves the way for similar AI-driven innovations in various industrial sectors, promising a future where smart technology and traditional manufacturing processes coexist in harmony.