AUTOMATING DETECTION OF PICK PLACE NOZZLE ANOMALIES (AN IIOT CASE STUDY)

摘要

Smart manufacturing requires quick decision. This means capabilities to monitor, perform data analysis, identify root cause and predict behaviors. The Digital Transformation has enabled Big Data solutions to capture real time data at a large scale from manufacturing equipment and systems. Tools exist to transform the data into meaningful insights but typically required users to observe the condition as it develops limiting its effectiveness. In this work, the initial steps of data acquisition, storage and processing will use the best of secured edge and cloud computing environments. The business value is to provide smart manufacturing in electronic assembly by adding business intelligence into decision-making. This creates actionable analytics by measuring and visualizing the SMT pick and place machine nozzle performance in real time. This allows operators and production support to "see" a nozzle anomaly in production. To detect these nozzle-level anomalies, an algorithm to track performance over time was developed. This algorithm is configurable to adjust its sensitivity for detecting an anomaly and notifying support personal. This algorithm is coupled with machine learning to forecast the performance of a specific nozzle. These tools will help identify anomalies, and trends for reducing downtime and defects while driving operations productivity.