유제품 산업의 품질검사를 위한 빅데이터 플랫폼 개발: 머신러닝 접근법 Building an Analytical Platform of Big Data for Quality Inspection in the Dairy Industry: A Machine Learning Approach원문보기
품질검사는 중간상품이나 최종상품을 품질관리 표준을 만족하는 양품과 불량품으로 분리하는 일을 수행한다. 대량생산체계에서 품질을 수작업으로 검사하는 것은 일관성과 효율성을 저하시키므로 대량으로 생산되는 상품의 품질을 검사하는 것은 다수의 공정에서 기계에 의한 자동 확인과 분류를 포함하게 된다. 생산공정에서 발생하는 데이터를 활용하여 공정을 개선하고 최적화하려는 선행 연구들이 많았음에도 불구하고, 실시간에 많은 데이터를 처리하는데 있어서의 기술적인 한계로 인해 실제 구현에서의 제약이 많이 있었다. 최근 빅데이터에 관한 연구에서는 데이터 처리기술을 개선하였고, 실시간에 데이터를 수집, 처리, 분석하는 과정을 가능하게 하게 하고 있다. 본 논문에서는 품질검사를 위한 빅데이터 적용의 단계와 세부사항을 제안하고, 유제품 산업에 적용 사례를 제시하려고 한다. 먼저 선행 연구들을 조사하고, 제조 부문에 적용할 수 있는 빅데이터 분석절차를 제안하며 제안된 방법의 실현가능성을 평가하기 위해서, 유제품 산업 분야의 품질검사과정 중 하나에 회선신경망(Convolutional Neural Network) 기술 및 랜덤포레스트(Random Forest) 기술을 적용하였다. 품질검사를 위해 제품의 뚜껑 및 빨대의 사진을 수집, 처리, 분석하여, 결함 여부를 판단하고, 과거 품질 검사결과와 비교하였다. 제안된 방법은 과거에 수행되었던 품질검사에 비해 분류 정확성 측면에서 의미 있는 개선을 확인할 수 있었다. 본 연구를 통해, 유제품 산업의 빅데이터 활용을 통한 품질검사 정확도 개선 가능성을 확인하였다.
품질검사는 중간상품이나 최종상품을 품질관리 표준을 만족하는 양품과 불량품으로 분리하는 일을 수행한다. 대량생산체계에서 품질을 수작업으로 검사하는 것은 일관성과 효율성을 저하시키므로 대량으로 생산되는 상품의 품질을 검사하는 것은 다수의 공정에서 기계에 의한 자동 확인과 분류를 포함하게 된다. 생산공정에서 발생하는 데이터를 활용하여 공정을 개선하고 최적화하려는 선행 연구들이 많았음에도 불구하고, 실시간에 많은 데이터를 처리하는데 있어서의 기술적인 한계로 인해 실제 구현에서의 제약이 많이 있었다. 최근 빅데이터에 관한 연구에서는 데이터 처리기술을 개선하였고, 실시간에 데이터를 수집, 처리, 분석하는 과정을 가능하게 하게 하고 있다. 본 논문에서는 품질검사를 위한 빅데이터 적용의 단계와 세부사항을 제안하고, 유제품 산업에 적용 사례를 제시하려고 한다. 먼저 선행 연구들을 조사하고, 제조 부문에 적용할 수 있는 빅데이터 분석절차를 제안하며 제안된 방법의 실현가능성을 평가하기 위해서, 유제품 산업 분야의 품질검사과정 중 하나에 회선신경망(Convolutional Neural Network) 기술 및 랜덤포레스트(Random Forest) 기술을 적용하였다. 품질검사를 위해 제품의 뚜껑 및 빨대의 사진을 수집, 처리, 분석하여, 결함 여부를 판단하고, 과거 품질 검사결과와 비교하였다. 제안된 방법은 과거에 수행되었던 품질검사에 비해 분류 정확성 측면에서 의미 있는 개선을 확인할 수 있었다. 본 연구를 통해, 유제품 산업의 빅데이터 활용을 통한 품질검사 정확도 개선 가능성을 확인하였다.
As one of the processes in the manufacturing industry, quality inspection inspects the intermediate products or final products to separate the good-quality goods that meet the quality management standard and the defective goods that do not. The manual inspection of quality in a mass production syste...
As one of the processes in the manufacturing industry, quality inspection inspects the intermediate products or final products to separate the good-quality goods that meet the quality management standard and the defective goods that do not. The manual inspection of quality in a mass production system may result in low consistency and efficiency. Therefore, the quality inspection of mass-produced products involves automatic checking and classifying by the machines in many processes. Although there are many preceding studies on improving or optimizing the process using the data generated in the production process, there have been many constraints with regard to actual implementation due to the technical limitations of processing a large volume of data in real time. The recent research studies on big data have improved the data processing technology and enabled collecting, processing, and analyzing process data in real time. This paper aims to propose the process and details of applying big data for quality inspection and examine the applicability of the proposed method to the dairy industry. We review the previous studies and propose a big data analysis procedure that is applicable to the manufacturing sector. To assess the feasibility of the proposed method, we applied two methods to one of the quality inspection processes in the dairy industry: convolutional neural network and random forest. We collected, processed, and analyzed the images of caps and straws in real time, and then determined whether the products were defective or not. The result confirmed that there was a drastic increase in classification accuracy compared to the quality inspection performed in the past.
As one of the processes in the manufacturing industry, quality inspection inspects the intermediate products or final products to separate the good-quality goods that meet the quality management standard and the defective goods that do not. The manual inspection of quality in a mass production system may result in low consistency and efficiency. Therefore, the quality inspection of mass-produced products involves automatic checking and classifying by the machines in many processes. Although there are many preceding studies on improving or optimizing the process using the data generated in the production process, there have been many constraints with regard to actual implementation due to the technical limitations of processing a large volume of data in real time. The recent research studies on big data have improved the data processing technology and enabled collecting, processing, and analyzing process data in real time. This paper aims to propose the process and details of applying big data for quality inspection and examine the applicability of the proposed method to the dairy industry. We review the previous studies and propose a big data analysis procedure that is applicable to the manufacturing sector. To assess the feasibility of the proposed method, we applied two methods to one of the quality inspection processes in the dairy industry: convolutional neural network and random forest. We collected, processed, and analyzed the images of caps and straws in real time, and then determined whether the products were defective or not. The result confirmed that there was a drastic increase in classification accuracy compared to the quality inspection performed in the past.
* AI 자동 식별 결과로 적합하지 않은 문장이 있을 수 있으니, 이용에 유의하시기 바랍니다.
제안 방법
Moreover, it is also necessary to study the result of improved classification accuracy in the aspect of ROI (Return On Investment). Another limitation is that this study focused only on implementing big data analytic system and applying the system to the inspection process. We need to develop new techniques customized for manufacturing processes.
For this study, “Real-time process improvement” was selected as the purpose of this analysis; the packaging step was selected for analysis since the data could be easily collected and the analysis result could be immediately applied to the process. At the time of this study, the quality was determined with a vision inspection system. The problem was that, although the system was well-equipped to identify good-quality products, it generated too many “false alarms” that considered good-quality products to be defective due to high sensitivity.
For the pilot test, the image data size was decreased to 8GB or less by reducing the image size from 640x480 to 320x240. The images were converted into numbers, segmented into training, validation, and test data, and then used for learning with CNN (Convolutional Neural Network) and RF (Random Forest) using various tuning values. Samples of cap images is illustrated in [Figure 5].
The main criteria for inspections include cap availability, print clarity of expiration date by the top image analysis, the location of the straw by the side image analysis.
Stojanovic proposed a conceptual model consisting of 6 layers for data-based quality management in a manufacturing process and suggested a method of detecting anomaly in real time. The study confirmed its feasibility by applying the analysis of big data collected in a microwave oven fan manufacturing process (Stojanovic et al., 2016). In another study, Stojanovic proposed a cluster analysis technique to identify unusualities among the real-time data for process improvement (Stojanovic et al.
This study proposed the key procedure for applying big data analysis to the manufacturing process and presented the result of applying the proposed procedure to the packaging process of dairy product manufacturing processes. The analysis result of vision data using two analysis techniques - CNN and RF - indicated that the new procedure significantly improved the classification accuracies of both cap and straw quality inspections compared to the past methods.
To investigate the feasibility of the analytical procedure presented in this study, a company was selected, and the procedure was applied to a manufacturing process.
성능/효과
To identify straw defect, the criterion related to attachment was divided into two types unlike the existing criterion. The analysis result indicated that CNN showed higher accuracy than RF in determining defects; thus, CNN was selected as the straw defect determination technique.
To identify cap defect, the criterion related to expiration date was divided into two types unlike the existing criteria. The analysis result indicated that RF showed higher accuracy than CNN in determining defects; thus, RF was selected as the cap defect determination technique.
This study proposed the key procedure for applying big data analysis to the manufacturing process and presented the result of applying the proposed procedure to the packaging process of dairy product manufacturing processes. The analysis result of vision data using two analysis techniques - CNN and RF - indicated that the new procedure significantly improved the classification accuracies of both cap and straw quality inspections compared to the past methods. RF showed better performance in the cap quality inspection, whereas CNN showed better performance in the straw quality inspection.
후속연구
The used method and process of analysis and application of this study would not expand its scope and methodologies to the general field of similar business cases theoretically. It did not confirm whether the proposed analysis technique was a universal method that could be applied to various quality inspections, and there was no appropriate analysis of the cost to benefit ratio as needed by the companies to decide on the application of big data analysis. To overcome such limitations, future studies should apply the big data analysis infrastructure built for this study to other processes, particularly to a study on power energy efficiency, which is actively being studied at present.
참고문헌 (15)
Chen, C.L. P., C.-Y. Zhang, "Data-intensive applications, challenges, techniques and technologies: A survey on Big Data", Information Sciences, Vol. 275 (2014), 314-347.
Ginsberg J., M. H. Mohebbi, R. S. Patel, L. Brammer, M. S. Smolinski, L. Brilliant, "Detecting influenza epidemics using search engine query data", Nature, Vol. 457(2000), 1012-1014.
Harding, J. A., M. Shahbaz, Srinivas, A. Kusiak, "Data Mining in Manufacturing: A Review", Journal of Manufacturing Science and Engineering, Vol. 128 (2006), 969-976.
Jagadish, H., J. Gehrke, A. Labrinidis, Y. Papakonstantinou, J. M. Patel, R. Ramakrishnan, C. Shahabi, "Big data and its technical challenges," Communications of the ACM, Vol. 57, No. 7 (2014), 86-94.
Lee, Y.-H., M.-S. Song, S.-J. Ha, T.-H. Baek, S.-Y. Son, "Big data Cloud Service for Manufacturing Process Analysis", The Korean journal of Bigdata, Vol.1, No.1 (2016), 41-51.
Li, Q., Z., Zhong, L. Zhengping, L., Yong, Rail Inspection Meets Big Data: Methods and Trends, Proceedings of 18th International Conference on Network-Based Information Systems (2015), 302-308.
MESA, "MESA International Survey", Technical Report, 2014.
Moyne, J., J. Samantaray, M. Armacost, "Big Data Capabilities Applied to Semiconductor Manufacturing Advanced Process Control", IEEE transactions on semiconductor manufacturing, Vol. 29, No. 4(2016), 283-291.
Kaisler S., F. Armour, J. A. Espinosa, W. M., "Big Data: Issues and Challenges Moving Forward", Proceedings of 2013 46th Hawaii International Conference on System Sciences (2013), 995-1004.
Kim, S. H., Jang, S. Y. and Yang, K. H. 2017. Analysis of the Determinants of Software-as-a-Service Adoption in Small Businesses: Risks, Benefits, and Organizational and Environmental Factors. Journal of Small Business Management, Vol. 55, No. 2 (2017), 303-325.
Shin S., S.H. Kim, K. B Song, "2015 Big Data Global Casebook", Technical Report, National Information Society Agency, 2015.
Hwang, J. S., Kim, S. H. and Lee, H. 2015. Breaking the Myths of the It Productivity Paradox. KSII Transactions on Internet and Information Systems, Vol. 9, No. 1 (2015), 466-482.
Stojanovic, L., D., Marko, S., Nenad, S. Ljiljana, "Big data process analytics for continuous process improvement in manufacturing", Proceedings of 2015 IEEE International Conference on Big Data (2015), 1398-1407.
Stojanovic, L., D., Marko, S., Nenad, S., Aleksandar, "Big-data-driven anomaly detection in industry (4.0): An approach and a case study", Proceedings of 2016 IEEE International Conference on Big Data (2016), 1647-1652.
Zarembski, A. M. "Some examples of big data in railroad engineering", Proceedings in 2014 IEEE International Conference on Big Data (2014), 96-102.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.