Supervised Learning

Definition

Supervised Learning is a machine learning method that learns the relationship between inputs and outputs using labeled data. Similar to how an experienced worker teaches a new employee by saying "this is a good product, and that is defective," the algorithm is trained by providing example data with correct answers marked.

Applications in Manufacturing

Quality Inspection Automation

• Defect Classification: Vision systems automatically detect defects after learning from thousands of normal/defective product images
• Welding Quality Assessment: Real-time quality judgment performed by learning from labeled good/bad weld images

Predictive Maintenance

• Predicts equipment failure timing by learning from historical failure data and normal operation data
• Early detection of abnormal signs by connecting sensor data with failure history

Process Optimization

• Derives optimal conditions by learning the relationship between process parameters (temperature, pressure) and product quality data

Key Points

Success Factor: Sufficient quantity of accurately labeled data is essential. In manufacturing sites, quality inspection records, sensor logs, and operator judgment history become training data.

Practical Application: A phased approach is effective—initially training models using expert judgments as labels, then gradually expanding the scope of automation.