Smart manufacturing systems are considered the next generation of manufacturing applications. One important goal of the smart manufacturing system is to rapidly detect and anticipate failures to reduce maintenance cost and minimize machine downtime. This often boils down to detecting anomalies within the sensor data acquired from the system which has different characteristics with respect to the operating point of the environment or machines, such as, the RPM of the motor. In this paper, we analyze four datasets from sensors deployed in manufacturing testbeds. We detect the level of defect for each sensor data leveraging deep learning techniques. We also evaluate the performance of several traditional and ML-based forecasting models for predicting the time series of sensor data. We show that careful selection of training data by aggregating multiple predictive RPM values is beneficial. Then, considering the sparse data from one kind of sensor, we perform transfer learning from a high data rate sensor to perform defect type classification. We release our manufacturing database corpus (4 datasets) and codes for anomaly detection and defect type classification for the community to build on it. Taken together, we show that predictive failure classification can be achieved, paving the way for predictive maintenance.

Smart systems such as data-driven machine health monitoring are emerging as a powerful technology for advanced manufacturing as a result of the availability of low-cost sensors, wireless communication, and advances in computational capabilities. Machine learning represents a critical element of manufacturing equipment health monitoring due to its ability to efficiently process large amounts of data and move beyond traditional rule-based maintenance methods. Predictive maintenance (PdM) has become increasingly popular in manufacturing. PdM can detect faults, determine root causes of operation anomalies, estimate the current health state of a system, and predict the future state and time when a component will fail in the absence of an intervention. One weakness of many past studies is the lack of run-to-failure data from an actual production environment. This paper presents run-to-failure data for the air compressor of an injection molding machine. A Long Short-Term Memory (LSTM) Recurrent Neural Network (RNN) is proposed to detect bearing faults in the air compressor. The model achieves a 97.4% of prediction accuracy (95.3% of overall accuracy). Experiments for machine state classification are also conducted and the classification performance compares favorably with conventional models.