Multi-sensor monitoring for in-situ defect detection and quality assurance in laser-directed energy deposition
Additive manufacturing (AM), specifically laser-directed energy deposition (LDED), has evolved rapidly as a pivotal technology in the realm of Industry 4.0, gaining significant momentum in both industry and academia. Despite its advancements, LDED still faces significant challenges in terms of quali...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis-Doctor of Philosophy |
| Language: | English |
| Published: |
Nanyang Technological University
2024
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/180195 |
| _version_ | 1826124905072033792 |
|---|---|
| author | Chen, Lequn |
| author2 | Moon Seung Ki |
| author_facet | Moon Seung Ki Chen, Lequn |
| author_sort | Chen, Lequn |
| collection | NTU |
| description | Additive manufacturing (AM), specifically laser-directed energy deposition (LDED), has evolved rapidly as a pivotal technology in the realm of Industry 4.0, gaining significant momentum in both industry and academia. Despite its advancements, LDED still faces significant challenges in terms of quality consistency, dimensional accuracy, and process repeatability. These challenges are especially prominent in industries such as aerospace, marine, automotive, and defence, where demand for highvalue components is increasing. The primary bottleneck to leveraging the full potential of LDED is the lack of an automated, real-time quality monitoring and control system during the production process. This results in a substantial reliance on manual, empirical offline process planning and parameter tuning, leading to increased labour costs and potential product quality compromises.
This research tackles a major challenge in LDED: the lack of an automated online monitoring system and the heavy manpower reliance in offline process planning. The objective is to develop an online, self-adaptive framework for intelligent LDED quality assurance, focusing on in-situ defect detection and adaptive quality enhancement. The proposed framework is structured around four key sub-objectives: Firstly, a multi-sensor in-situ process monitoring platform is developed. This platform integrates data from various sensors, including an acoustic sensor, a shortwave infrared (SWIR) thermal camera, a laser displacement sensor, and a coaxial visible spectrum camera, alongside robot tool-centre-point (TCP) motion data. This multimodal dataset provides a comprehensive real-time view of the LDED process, which is the pre-requisite for in-situ defect detection. Secondly, a multimodal spatiotemporal data fusion and feature correlation analysis is conducted. This research involves synchronizing and registering the multimodal dataset within the 3D volumetric domain of the as-printed component. This step aims to quantitatively analyse cross-modality correlations between various data characteristics, such as acoustic signatures, melt pool visual characteristics, and temperature distributions. Understanding these correlations is the key to establishing effective machine learning (ML) models for defect detection. Thirdly, a location-dependent quality prediction model is developed using the fused dataset. Leveraging the fused multimodal dataset, this research aims to predict various defect occurrences and process anomalies (e.g., keyhole pores, cracks, laser-off, etc.) at different locations within the component. A comprehensive, data-driven ML model is developed for in-situ identification of potential defects during the LDED process. Fourthly, an adaptive quality enhancement strategy is proposed in LDED. The framework includes strategies for in-process quality enhancement based on feedback from location-specific quality predictions in the previous step. This involves decision-making regarding the subtractive removal of defective regions or additive repair of surface defects, enhancing the overall quality and dimension accuracy of the LDED-produced parts.
Furthermore, a unified software platform is developed to embody the online self-adaptation
framework. The proposed multimodal monitoring approach has achieved a significantly higher quality
prediction accuracy (96%) compared to traditional single-sensor-based defect identification. The
proposed MFDT framework lays the foundation for a self-adaptation LDED process, which allows for
in-process quality assurance and has the potential to achieve zero-defect autonomous AM processes.
The proposed approach not only facilitates broader industrial adoption of metal AM but also promotes
greater efficiency, reduced waste, and cleaner production practices. |
| first_indexed | 2024-10-01T06:27:59Z |
| format | Thesis-Doctor of Philosophy |
| id | ntu-10356/180195 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2025-03-09T13:53:40Z |
| publishDate | 2024 |
| publisher | Nanyang Technological University |
| record_format | dspace |
| spelling | ntu-10356/1801952024-10-07T01:58:13Z Multi-sensor monitoring for in-situ defect detection and quality assurance in laser-directed energy deposition Chen, Lequn Moon Seung Ki School of Mechanical and Aerospace Engineering Singapore Centre for 3D Printing Singapore Institute of Manufacturing Technology Advanced Remanufacturing and Technology Centre (ARTC) Chew Youxiang Liu Kui Yao Xiling skmoon@ntu.edu.sg Computer and Information Science Engineering Additive manufacturing Audio signal processing Machine learning Multimodal dta fusion Laser directed energy deposition Additive manufacturing (AM), specifically laser-directed energy deposition (LDED), has evolved rapidly as a pivotal technology in the realm of Industry 4.0, gaining significant momentum in both industry and academia. Despite its advancements, LDED still faces significant challenges in terms of quality consistency, dimensional accuracy, and process repeatability. These challenges are especially prominent in industries such as aerospace, marine, automotive, and defence, where demand for highvalue components is increasing. The primary bottleneck to leveraging the full potential of LDED is the lack of an automated, real-time quality monitoring and control system during the production process. This results in a substantial reliance on manual, empirical offline process planning and parameter tuning, leading to increased labour costs and potential product quality compromises. This research tackles a major challenge in LDED: the lack of an automated online monitoring system and the heavy manpower reliance in offline process planning. The objective is to develop an online, self-adaptive framework for intelligent LDED quality assurance, focusing on in-situ defect detection and adaptive quality enhancement. The proposed framework is structured around four key sub-objectives: Firstly, a multi-sensor in-situ process monitoring platform is developed. This platform integrates data from various sensors, including an acoustic sensor, a shortwave infrared (SWIR) thermal camera, a laser displacement sensor, and a coaxial visible spectrum camera, alongside robot tool-centre-point (TCP) motion data. This multimodal dataset provides a comprehensive real-time view of the LDED process, which is the pre-requisite for in-situ defect detection. Secondly, a multimodal spatiotemporal data fusion and feature correlation analysis is conducted. This research involves synchronizing and registering the multimodal dataset within the 3D volumetric domain of the as-printed component. This step aims to quantitatively analyse cross-modality correlations between various data characteristics, such as acoustic signatures, melt pool visual characteristics, and temperature distributions. Understanding these correlations is the key to establishing effective machine learning (ML) models for defect detection. Thirdly, a location-dependent quality prediction model is developed using the fused dataset. Leveraging the fused multimodal dataset, this research aims to predict various defect occurrences and process anomalies (e.g., keyhole pores, cracks, laser-off, etc.) at different locations within the component. A comprehensive, data-driven ML model is developed for in-situ identification of potential defects during the LDED process. Fourthly, an adaptive quality enhancement strategy is proposed in LDED. The framework includes strategies for in-process quality enhancement based on feedback from location-specific quality predictions in the previous step. This involves decision-making regarding the subtractive removal of defective regions or additive repair of surface defects, enhancing the overall quality and dimension accuracy of the LDED-produced parts. Furthermore, a unified software platform is developed to embody the online self-adaptation framework. The proposed multimodal monitoring approach has achieved a significantly higher quality prediction accuracy (96%) compared to traditional single-sensor-based defect identification. The proposed MFDT framework lays the foundation for a self-adaptation LDED process, which allows for in-process quality assurance and has the potential to achieve zero-defect autonomous AM processes. The proposed approach not only facilitates broader industrial adoption of metal AM but also promotes greater efficiency, reduced waste, and cleaner production practices. Doctor of Philosophy 2024-09-23T08:36:34Z 2024-09-23T08:36:34Z 2024 Thesis-Doctor of Philosophy Chen, L. (2024). Multi-sensor monitoring for in-situ defect detection and quality assurance in laser-directed energy deposition. Doctoral thesis, Nanyang Technological University, Singapore. https://hdl.handle.net/10356/180195 https://hdl.handle.net/10356/180195 10.32657/10356/180195 en Career Development Fund (Grant No. C210812030) Singapore Centre for 3D Printing (SC3DP), the National Research Foundation, Prime Minister’s Office, Singapore under its Medium-Sized Centre funding scheme Agency for Science, Technology and Research (A*STAR) of Singapore through RIE2025 MTC IAF-PP grant (Grant No. M22K5a0045) This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0). application/pdf Nanyang Technological University |
| spellingShingle | Computer and Information Science Engineering Additive manufacturing Audio signal processing Machine learning Multimodal dta fusion Laser directed energy deposition Chen, Lequn Multi-sensor monitoring for in-situ defect detection and quality assurance in laser-directed energy deposition |
| title | Multi-sensor monitoring for in-situ defect detection and quality assurance in laser-directed energy deposition |
| title_full | Multi-sensor monitoring for in-situ defect detection and quality assurance in laser-directed energy deposition |
| title_fullStr | Multi-sensor monitoring for in-situ defect detection and quality assurance in laser-directed energy deposition |
| title_full_unstemmed | Multi-sensor monitoring for in-situ defect detection and quality assurance in laser-directed energy deposition |
| title_short | Multi-sensor monitoring for in-situ defect detection and quality assurance in laser-directed energy deposition |
| title_sort | multi sensor monitoring for in situ defect detection and quality assurance in laser directed energy deposition |
| topic | Computer and Information Science Engineering Additive manufacturing Audio signal processing Machine learning Multimodal dta fusion Laser directed energy deposition |
| url | https://hdl.handle.net/10356/180195 |
| work_keys_str_mv | AT chenlequn multisensormonitoringforinsitudefectdetectionandqualityassuranceinlaserdirectedenergydeposition |