In situ identification of laser directed energy deposition condition based on acoustic emission

Additive manufacturing with remarkable merits has enormous potential and application prospects in plenty of fields. Unfortunately, the inconsistent quality of specimens fabricated is severely hindering this technology from meeting the criteria for commercialization. Exploitation of an efficient process monitoring method is essential to evaluate the usability of fabricated parts. Therefore, an in situ monitoring system based on acoustic emission sensing was proposed to acquire the information (elastic waves generated inside parts) during the manufacturing process. Initially, time–frequency analysis of the captured signals in the operation conditions of powder feeding solely, laser melting solely and deposition was carried out. The analysis had contributions to confirm the frequency band largely indicating powder colliding with substrate. Then, twelve typical statistical and three specific features from time and frequency domain were extracted to assist further information mining. Clustering analysis based on k-means was implemented to reveal the cohesiveness and isolation of samples from different classes in the feature space. Support vector machine, random forest and back propagation neural network were employed to construct models to identify different operation conditions or all groups of tracks. The results showed that all models were qualified to identify the operation conditions with a high precision. In the recognition of all group of tracks, support vector machine model achieved the best performance with an accuracy of 94 %, which definitely indicates the feasibility and potential of AE sensing in the monitoring of laser directed energy deposition.