With the gradual popularization of IoT and 5G technologies, the technical threshold for manufacturers to use sensors to collect the real-time status of equipment has been greatly reduced. Fault detection and classification are important functions for advanced process control. Traditionally, domain engineers use domain knowledge and experience to find suitable key intervals and convert them into indicators for process monitoring. This method consumes a lot of manpower and time, and under the pressure of time, the feasible solution is usually found rather than the best solution. This technology can deal with a huge amount of profile data collected from the device sensors in equipment with quality measurement, through ensemble learning and machine learning algorithms, to perform automatic feature extraction and definition, providing engineers the ranking of features for decision support.

In the past, when using equipment sensing data for fault detection and classification, it is necessary to rely on expert domain knowledge to define appropriate key intervals and statistics in advance for analysis and monitoring. Now, when the quality measurement is feedbacked, this technology precisely defines the key intervals and feature statistics, and provides decision indicators as the confidence level of the model, reducing the engineer’s time expenditure and closing the experience gap, and accelerating the mass production of products.

This technology can be applied to process control, especially for fault detection and classification. Since the technology does not rely on domain knowledge, it can be quickly applied to various industries with various types of analytical needs, such as the semiconductor industry, electronics industry, or the aviation manufacturing industry.