Study Of The Monitoring Technique And Its Application For Additive Manufacturing Based On Acoustic Emission

The additive manufacturing (AM) technologies have been rapidly evolved in recent years,and the importances to ensure process reliability and overall product quality of AM have been well addressed in both industry and academia. The AM products are built by adding material in a layer-by-layer scheme, which is different from traditional subtractive manufacturing. Process monitoring is needed in AM because many material phase changes are involved in AM process, product de-fects and abnormal states may also occur during long fabrication process, especially in fabricating complex and large products. The material defects and abnormal states in AM will stimulate elastic waves which are closely related to acoustic emission (AE) phenomena, but this nesearch field is still in its primary stage. Thus,in-depth and systematic studies on the AE monitoring strategy,signal acquisition, feature extraction, related state identification methods and their applications in AM are also required. In this study, the AE is adopted as the nondestructive sensing technique,and data-driven method based on sensory information is used to study the monitoring of products,machine operation states, and the related scientific issues in AM. The overall obictives of this study include revealing the associations among AE signals and material properties, typical defects,abnormal states, etc., in AM processes; studying the specific and efficient AE feature extraction methods; developing the decision making and identification methods based on typical machine learning algorithms for defects or abnormal states; and put forward a R&D method of a multichan-nel AE monitoring prototype system for AM processes. The specific contributions of this study are introduced as follows:Chapter1 The introductions of the scientific backgrounds and goals of this study are pro-vided. The state-of-art of AM, process monitoring, and AE technique are reviewed. The recent research progresses and the key issues in monitoring AM processes are discussed, and the major contents of this study are also introduced.Chapter 2 Aiming at the ferrous material surface burn issue in typical laser sintering AM process, preliminary study on AE monitoring strategy, feature extraction, and analysis method is conducted. A feature extraction and de-noising method based on EEMD is developed. The AE monitoring experiment platform is set up that a pulse laser machine is used to induce surface burn in the experimental study. Raw AE signal data are acquired from experiments. The associations among raw AE signal features, material properties, ferrous material surface burn phenomena, etc.,are investigated, and the key IMFs that are closely related to surface burn are extracted for the purpose of surface burn monitoring.Chapter 3 Aiming at the typical product defects in AM, the major patterns and influence factors of the defects of FDM products,such as material debonding, peeling off, rubbing, and scratching are analyzed, and studies on the monitoring method are carried out. An AE monitoring method based on AE hit features and unsupervised SOM algorithm is put forward, and this method is employed and tested in experimental study. In addition, the major frequency bands of AE hits of the typical defects, and the evolution of each defect mode are studied.Chapter 4 Aiming at the AM machine operation and related abnormal states, studies on the AE-based online monitoring method are carried out. Several normal and abnormal states of FDM print head, such as material run out, filament breakage, and blockage, are focused. AE monitoring methods based on SVM and K-means are developed, where the segmental analysis is used to further compress the AE hit feature data. The states of print head can be identified in experimental studies.Chapter 5 Aiming at the AE online monitoring method's performance in Chapter 4, an op-timized method is put forward to improve the adaptivity and real-time performance by adopting PCA and HSMM algorithms at the same time. The optimized method can select and compress AE hit features adaptively, and the real-time performance of state identification is improved according to the experiment results.Chapter 6 Based on the above study, the R& D method of a multichannel AE monitoring prototype system for AM is studied. A case study for FDM, including demand analysis, hard-ware design, component selection, workflow and software design, system integration, and etc.,is conducted. This study provides technical support for realizing intelligent monitoring in AM and developing related monitoring system products.