A convolutional approach to quality monitoring for laser manufacturing

Gonzalez-Val, Carlos; Pallas, Adrian; Panadeiro, Verónica; Rodríguez, Almudena Suárez

doi:10.1007/s10845-019-01495-8

Cited by 70 publications

(23 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Thermal images and convolutional neural networks work well in combination, as shown by Gonzales-Val et al [31]. They proposed a CNN architecture to predict dilution in laser metal deposition as well as defects in laser welding based on infrared images.…”

Section: Introductionmentioning

confidence: 98%

Deep Learning and Conventional Machine Learning for Image-Based in-Situ Fault Detection During Laser Welding: A Comparative Study

Knaak¹,

Kröger²,

Schulze³

et al. 2021

Preprint

View full text Add to dashboard Cite

An effective process monitoring strategy is a requirement for meeting the challenges posed by increasingly complex products and manufacturing processes. To address these needs, this study investigates a comprehensive scheme based on classical machine learning methods, deep learning algorithms, and feature extraction and selection techniques. In a first step, a novel deep learning architecture based on convolutional neural networks (CNN) and gated recurrent units (GRU) is introduced to predict the local weld quality based on mid-wave infrared (MWIR) and near-infrared (NIR) image data. The developed technology is used to discover critical welding defects including lack of fusion (false friends), sagging and lack of penetration, and geometric deviations of the weld seam. Additional work is conducted to investigate the significance of various geometrical, statistical, and spatio-temporal features extracted from the keyhole and weld pool regions. Furthermore, the performance of the proposed deep learning architecture is compared to that of classical supervised machine learning algorithms, such as multi-layer perceptron (MLP), logistic regression (LogReg), support vector machines (SVM), decision trees (DT), random forest (RF) and k-Nearest Neighbors (kNN). Optimal hyperparameters for each algorithm are determined by an extensive grid search. Ultimately, the three best classification models are combined into an ensemble classifier that yields the highest detection rates and achieves the most robust estimation of welding defects among all classifiers studied, which is validated on previously unknown welding trials.

show abstract

Section: Introductionmentioning

confidence: 98%

Deep Learning and Conventional Machine Learning for Image-Based in-Situ Fault Detection During Laser Welding: A Comparative Study

Knaak¹,

Kröger²,

Schulze³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Neural networks have also been used to monitor image based processes using a style of network called convolutional neural networks (CNNs). Gonzalez-Val et al (2020) performed experiments in laser metal deposition and welding to determine machining quality via two different metrics. They determined that it would be possible to create in in line defect detection system, allowing for early correction of errors.…”

Section: Introductionmentioning

confidence: 99%

Machine learning for multi-dimensional optimisation and predictive visualisation of laser machining

et al. 2021

View full text Add to dashboard Cite

Interactions between light and matter during short-pulse laser materials processing are highly nonlinear, and hence acutely sensitive to laser parameters such as the pulse energy, repetition rate, and number of pulses used. Due to this complexity, simulation approaches based on calculation of the underlying physical principles can often only provide a qualitative understanding of the inter-relationships between these parameters. An alternative approach such as parameter optimisation, often requires a systematic and hence time-consuming experimental exploration over the available parameter space. Here, we apply neural networks for parameter optimisation and for predictive visualisation of expected outcomes in laser surface texturing with blind vias for tribology control applications. Critically, this method greatly reduces the amount of experimental laser machining data that is needed and associated development time, without negatively impacting accuracy or performance. The techniques presented here could be applied in a wide range of fields and have the potential to significantly reduce the time, and the costs associated with laser process optimisation.

show abstract

“…The use of CNNs in the video processing field, in applications useful to industry, is still significant and growing [ 17 ]. CNN approaches are capable of monitoring laser-based manufacturing processes [ 17 , 18 , 19 , 20 ]. However, it is difficult to find solutions that meet the automated assessment and measurement requirements of the activity of the single worker.…”

Section: Introductionmentioning

confidence: 99%

A Deep Learning-Based Model for the Automated Assessment of the Activity of a Single Worker

Patalas‐Maliszewska

Halikowski

2020

Sensors

View full text Add to dashboard Cite

Nowadays, it is necessary to verify the accuracy of servicing work, undertaken by new employees, within a manufacturing company. A gap in the research has been observed in effective methods to automatically evaluate the work of a newly employed worker. The main purpose of the study is to build a new, deep learning model, in order to automatically assess the activity of the single worker. The proposed approach integrates the methods known as CNN, CNN + SVM, CNN + R-CNN, four new algorithms and a piece of work from a selected company, using this as an own-created dataset, in order to create a solution enabling assessment of the activity of single workers. Data were collected from an operational manufacturing cell without any guided or scripted work. The results reveal that the model developed is able to accurately detect the correctness of the work process. The model’s accuracy mostly exceeds current state-of-the-art methods for detecting work activities in manufacturing. The proposed two-stage approach, firstly, assigning the appropriate graphic instruction to a given employee’s activity using CNN and then using R-CNN to isolate the object from the reference frames, yields 94.01% and 73.15% accuracy of identification, respectively.

show abstract

A convolutional approach to quality monitoring for laser manufacturing

Cited by 70 publications

References 21 publications

Deep Learning and Conventional Machine Learning for Image-Based in-Situ Fault Detection During Laser Welding: A Comparative Study

Deep Learning and Conventional Machine Learning for Image-Based in-Situ Fault Detection During Laser Welding: A Comparative Study

Machine learning for multi-dimensional optimisation and predictive visualisation of laser machining

A Deep Learning-Based Model for the Automated Assessment of the Activity of a Single Worker

Contact Info

Product

Resources

About