Machine-learning-based quality-level-estimation system for inspecting steel microstructures

Abstract

Quality control of special steel is accomplished through visual inspection of its microstructure based on microscopic images. This study proposes an “automatic-quality-level-estimation system” based on machine learning. Visual inspection of this type is sensory-based, so training data may include variations in judgments and training errors due to individual differences between inspectors, which makes it easy for a drop in generalization performance to occur due to overfitting. To deal with this issue, we here propose the preprocessing of inspection images and a data augmentation technique. Preprocessing reduces variation in images by extracting features that are highly related to the level of quality from inspection images. Data augmentation, meanwhile, suppresses the problem of overfitting when training with a small number of images by taking into account information on variation in judgment values obtained from on-site experience. While the correct-answer rate for judging quality level by an inspector was about 90%, the proposed method achieved a correct-answer rate of 92.5%, which indicates that the method shows promise for practical application.

Sistemas de visão aplicados em processos de qualidade automotiva: Uma revisão sistemática

“…Although the loss value for the training data is generally smaller than the loss value for the validation data in general learning, the proposed method showed a significantly lower loss value for the validation data. This discrepancy can be attributed to dispersion of correct answer values caused by data expansion, suggesting the potential to further improve judgment accuracy by adjusting the distribution of variance of correct answer values [3].Overall, the experimental results confirmed the effectiveness of the proposed method, demonstrating an improvement in estimation accuracy and generalization performance compared to the comparative method [3].…”

“…The experiment used a four-part crossvalidation approach with a training-validation-testing ratio of 6.0:1.5:2.5. A comparative method without data expansion was also evaluated for independent verification [3]. As can be seen at Figure 5.…”

“…This result indicates that data expansion improved performance. The proposed method achieved a correct response rate equal to that of a human inspector, which was considered the performance target for practical use [3]. The learning curve in Figure 6 illustrates the value of the loss function during training.…”

“…Implementing the system will eliminate the influence of personal differences between inspectors and reduce inspection costs. In addition, the application of this system in the inspection of metals other than alloy tool steel verified in this study will be investigated, aiming to expand the scope of automation based on this system [3].…”

Applications of Deep Learning for Drug Discovery Systems with BigData

“…Therefore, it is prone to overfitting, which is a condition where the learning data are overfitted, generalization is not possible, and high accuracy cannot be achieved with unknown data. To suppress overfitting and achieve high accuracy, a large amount of training data is required, but in practice, it is not always easy to collect a large amount of training data [162][163][164][165][166][167][168][169][170][171][172][173][174][175][176][177][178]. Moreover, it takes a long time to input a large amount of training data into a large-scale DNN and execute the learning process.…”