Evaluating AI Algorithms for Identifying Anomalies in Composite Additive Manufacturing

IF 2.9 4区材料科学 Q3 MATERIALS SCIENCE, COMPOSITES

Applied Composite Materials Pub Date : 2025-05-08 DOI:10.1007/s10443-025-10340-6

Deepak Kumar, Yongxin Liu, Sirish Namilae

引用次数: 0

Abstract

Despite significant progress in additive manufacturing, processing defects remain a persistent challenge. Artificial intelligence (AI) enabled early defect detection and process optimization is promising solution for this problem. In this study, real image data from a composite 3D printing setup was used to evaluate the anomaly detection performance of three models: Autoencoder, Support Vector Machine (SVM), and the Zero Bias Deep Neural Network (DNN). The results demonstrate that the Zero bias model achieved an accuracy of 97.96%, significantly outperforming the Autoencoder (93.38%) and SVM (89.80%). Multiple thresholds in the zero bias model enable explain ability.

查看原文本刊更多论文

评估人工智能算法在复合材料增材制造中的异常识别

尽管增材制造取得了重大进展，但加工缺陷仍然是一个持续的挑战。人工智能（AI）支持早期缺陷检测和过程优化是解决这一问题的有希望的解决方案。在这项研究中，使用来自复合3D打印装置的真实图像数据来评估三种模型的异常检测性能：自动编码器，支持向量机（SVM）和零偏差深度神经网络（DNN）。结果表明，零偏差模型的准确率达到97.96%，显著优于Autoencoder（93.38%）和SVM（89.80%）。零偏差模型中的多个阈值使解释能力得以实现。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Applied Composite Materials 工程技术-材料科学：复合

CiteScore

4.20

自引率

4.30%

发文量

审稿时长

1.6 months

期刊介绍： Applied Composite Materials is an international journal dedicated to the publication of original full-length papers, review articles and short communications of the highest quality that advance the development and application of engineering composite materials. Its articles identify problems that limit the performance and reliability of the composite material and composite part; and propose solutions that lead to innovation in design and the successful exploitation and commercialization of composite materials across the widest spectrum of engineering uses. The main focus is on the quantitative descriptions of material systems and processing routes. Coverage includes management of time-dependent changes in microscopic and macroscopic structure and its exploitation from the material''s conception through to its eventual obsolescence.