Application plan

• Aiming to create more accurate process designs through AI-based big data analysis to reduce the possibility of failure and standardize optimal process designs.
• To derive a specified result value (hardness), find the X factor factor (pressure rate, annealing condition) that affects this result value and apply it to the process design
• Based on 10 years of manufacturing data, the AI solution model is trained to find the rolling rate and annealing conditions, and then the accuracy is tested by adding raw data and comparing them.

AI analysis methodology

• A regression analysis technique was used, and the independent variables were composed of material steel type, hardness specification, hardness, and product thickness, and the dependent variables were annealing conditions and pressure drop rate.
• Use the Multinomial Logistic Regression Learning Method with Softmax Activation Function in the AI Model for Predicting Eradication Conditions, Categorical Data
• Use the basic linear regression learning method by adding annealing conditions as an independent variable to predict the pressure drop rate.

AI Field Application
Results

• Enhancing competitiveness through data-driven manufacturing innovation
• Use manufacturing data and AI analysis experience to support field data-based decision making to address line or process issues
• Reduce risk, fatigue, and time by systematizing process designs that were focused on some workers

Application plan

• After data collection, apply preprocessing to ensure high-quality learning data
• Database data and apply CNN techniques to secure large amounts of data for AI learning
• Distributed storage of learning data using a cloud environment to efficiently manage large-capacity learning data
• Saving image data (atypical data) through a vision inspector using a NoSQL-based data store

AI analysis methodology

• This AI solution utilizes the CNN algorithm technique
• CNNs are especially useful in finding patterns to recognize images
• Learn directly from data and categorize images using patterns, eliminating the need to manually extract features
• Because of these advantages, it is widely used in fields that require object recognition or computer vision, such as self-driving cars, face recognition, and good/defective products using images.

AI Field Application
Results

• Automation of inspection: Integration with vision system enables automation of inspection of classification of good and defective products and classification of defects.
• Improvement of customer claim rate: Expectation of improvement of customer claim rate by selecting defective products through product inspection and shipping products
• Product total inspection through AI-based system: Improvement of claim rate through AI total inspection. Expect to reduce claim processing costs due to this
• Cost reduction: Lower costs of disposing of good products at the same time as blocking bad delivery by accurately classifying ambiguous quantities/defects

Application plan

• Aiming to create more accurate process designs through AI-based big data analysis to reduce the possibility of failure and standardize optimal process designs.
• To derive a specified result value (hardness), find the X factor factor (pressure rate, annealing condition) that affects this result value and apply it to the process design
• Based on 10 years of manufacturing data, the AI solution model is trained to find the rolling rate and annealing conditions, and then the accuracy is tested by adding raw data and comparing them.

AI analysis methodology

1 LEFT/RIGHT Line Model Learning with Segmentation Model Using CNN
• Width and Height Threshold filtering of the LEFT and RIGHT lines
2 This AI solution utilizes supervised learning techniques
• Already knows the value of the outer diameter of the label and the result of its tolerance
• It is important to find optimal values such as suppliers, die life cycles, and coolant temperature that affect the label.
• Analyze the X factor when the external diameter value and the actual measured value deviating from the tolerance are collected

AI Field Application
Results

• This solution can prevent a large number of defects due to specification out in advance by providing an abnormal alarm system based on real-time measurement information that was not possible in the existing manual operation.
• You can guarantee customer product reliability.
• By incorporating this solution into the wire fresh field of various products, a low-cost AI-based fresh line abnormality alarm system can be utilized.