Search for: All records

Abstract Digital transformation of manufacturing industry, Smart Manufacturing, leverages continuous measurement of machines on the shop floor to make effective decisions and improve productivity metrics such as machine uptime and overall equipment efficiency (OEE). However, despite the declining sensor cost, the initial financial and technological skill requirements of digital transformation pose significant barriers for the overwhelming majority (90%) of the manufacturers who are classed as small and medium enterprises (SMEs). To lower this barrier, here we demonstrate an inexpensive (~ $40 per machine), data-efficient solution that extracts part-level productivity metrics of a CNC machine from its total current consumption alone. We introduce the concept of a part’s “fingerprint” and develop a set of methods that allows one to extract the fingerprints and utilize them to monitor each individual manufactured part and their cycle times. Testing on actual production data of over 3 three months in a part-counting task, the algorithms show a good match (96.2% overall accuracy) with manually logged production data is achieved. The presented fingerprint framework is general: it can be extended to multi-sensors, and multi-modal analytics. We expect that such a simple, yet cost-effective, solution will be accessible for a wide range of discrete manufacturers, facilitating the beginning of their digital transformation journey. 

Abstract This study introduces a non-invasive approach to monitor operation and productivity of a legacy pipe bending machine in real-time based on a lightweight convolutional neural network (CNN) model and internal sound as input data. Various sensors were deployed to determine the optimal sensor type and placement, and labels for training and testing the CNN model were generated through the meticulous collection of sound data in conjunction with webcam videos. The CNN model, which was optimized through hyperparameter tuning via grid search and utilized feature extraction using Log-Mel spectrogram, demonstrated notable prediction accuracies in the test. However, when applied in a real-world manufacturing scenario, the model encountered a significant number of errors in predicting productivity. To navigate through this challenge and enhance the predictive accuracy of the system, a buffer algorithm using the inferences of CNN models was proposed. This algorithm employs a queuing method for continuous sound monitoring securing robust predictions, refines the interpretation of the CNN model inferences, and enhances prediction outcomes in actual implementation where accuracy of monitoring productivity information is crucial. The proposed lightweight CNN model alongside the buffer algorithm was successfully deployed on an edge computer, enabling real-time remote monitoring. 

In this research we introduce the application of an optical fiber Fabry-Pérot interferometer in smart manufacturing. We used an optical fiber Fabry-Pérot interferometer to measure the distance between a moving target and a fixed optical fiber. When the target moves, the distance between the fiber and the target can be precisely determined. First, we monitored the distance between a fixed fiber and the surface of a rotating tool. By measuring the distance, we reconstructed the three-dimensional (3D) profile of the tool. We also introduce the method to calculate the runout and tool wear. To further improve the speed of this method, we developed machine learning models to find out the distance from the spectrum of the interferometer since the spectrum analyzing method is relatively slow. It was found that the Deep Neural Network model predicts the distance between the fiber and the target surface with a sufficient precision (< 4 μm) when measuring the straightness error of a computer numerical control (CNC) machine tool. The proposed method provides possibilities for noncontact precise monitoring especially in a limited space.