The Possibility of Applying Deep Learning Techniques in Product Quality Control / Case Study

Abstract

This study addresses the production status, defect resolution, and quality control issues in the soap department of Al-Maamoun Factory, part of the General Company for Food Products, using deep learning algorithms and techniques. The research problem was identified as defective production and the poor performance of the quality control department, which lacks modern technologies. To select the quality control criteria for liquid soap, the focus was on active ingredient (NaCl), viscosity, pH, and salts (AD). MATLAB - Ver 2024 was used to apply five algorithms: decision tree, Naive Bayes classifier, convolutional neural network, artificial neural network, and logistic model, along with SPSS-Ver.28. The results showed that out of 206 randomly selected bottles of liquid soap, 155 were compliant with specifications and 51 were non-compliant, constituting 25% of the sample. The study revealed that the decision tree algorithm is the best in identifying defect causes and classification accuracy. The objectives of this study include improving production quality using deep learning techniques, accurately identifying defect causes, reducing waste and costs, building an expert system for automated quality control, and raising awareness among officials about modern technologies. The study is significant for enriching scientific knowledge in this field, recognizing the importance of deep learning techniques in reducing defects, and adopting modern techniques to reduce inspection time and effort.