Artificial Intelligence in Manufacturing Industry: Product Quality Control System
Developing an AI-based alerting system to address the issue of maintaining the optimum size of plywood sheets during the manufacturing process.
THE CUSTOMER
The customer is one of the largest manufacturer, seller, and exporter of plywood, laminates, doors, PVCs, and veneers in India. With more than 5000+ employees and spread into a 6.6-acre area, the company’s plant is located at eastern part of India.
BUSINESS GOALS
The customer is a leading manufacturer and supplier of plywood with 25% market share in the organised plywood market of India. The brand value plays a vital role in their sales grwoth
Its revenue from operations was Rs 3,000 crore in 2021-22. In recent times, the company encountered some quality issues with their plywood products. According to their internal report, over 2% of their products are having quality issue which are untraceable by manual check points. The distributors are raising this concern for the last couple of years.
The faulty 2% lot is having the core sheet density issue. The core sheet in the ply holds the engineered wood firmly and makes it stronger. The optimal density of the core sheet is unevenly distributed which eventually affects the final product and make it weak. Manual checkpoints are unable to check the issue in the large set of manufactured products. Therefore, the customer has decided to adopt AI technologies to automate the product quality checking system with higher accuracy rate.
THE SOLUTION
After carefully assessing customer’s requirements, ThirdEye has designed an AI solution based on OpenCV models to detect the sheet size & density and provide real-time alerts to operators.
The proposed solution involves various AI-based techniques that enables the customer to minimize production waste and improve its bottom line. The solution involves the installation of a camera system above the production machine that captures images of the plywood sheets as they are inserted into the production machine. The images are processed using computer vision techniques to accurately measure the size and density of the core sheet. If the core sheet size is less than the specified breadth, or the density falls below the optimal measurement, the system trigger an alert, such as a visual indication with a red light, to notify the operator.
ThirdEye has developed an AI Image Processing and Anomaly Detection system with features like collecting real-rime data (image format), instant processing, object detection, and notification. The system was powered by AI techniques, backed by object detection technologies and complex algorithms. Our system drives the data-driven decision making program with real-time reporting to achieve the goals for the customer.
The process steps were as follows:
- Capture Real-time Image: Used high-resolution camera to capture a real-time image of the core sheets being inserted into the production machines. This was done using a webcam and other suitable camera device.
- Image Preprocessing: Applied necessary preprocessing techniques to enhance the image quality and made it suitable for measurement. These techniques included resizing, noise removal, and image enhancement.
- Reference Object Selection: Chose a reference object of known dimensions that could be easily identified in the image. That reference object placed near the core sheet for analytical purposes.
- Detect Reference Object: Utilized OpenCV’s image processing functions, such as edge detection or color segmentation, to locate and isolate the reference object in the image.
- Calculate Pixels per Metric Ratio: Measured the dimensions (width or height) of the reference object in pixels. Divided the known dimensions of the reference object by the measured dimensions to obtain the pixels per metric ratio.
- Detect Core Sheet: Applied suitable image processing techniques to locate and isolate the core sheet in the image. This was done by analyzing color, texture, or shape characteristics.
- Measure Core Sheet Length: Determined the length of the core sheet by measuring the dimensions of the detected core sheet in pixels.
- Convert to Real-world Units: Multiplied the measured length of the plywood sheet (in pixels) by the pixels per metric ratio obtained in step 5 to convert it to real-world units (e.g., millimeters, inches).
- Check Length Threshold: Compared the measured length of the plywood sheet with the desired threshold (e.g., 12 feet). When the length was less than the threshold, generated a real-time notification to alert the system.
- Alert by Lighting up a Red Bulb: Used a simple circuit connected to the system, light up a red light to indicate the error in the total length of the core sheet being inserted into the production system.
Technologies Incorporated:
- High Resolution Cameras
- TensorFlow
- Model Building, deploying, object detection and serving
- Pytorch
- OpenCV
- On-Prem Server
- RCNN (Region-based Convolutional Neural Network) Model
- Custom AI Algorithm
- PowerBI
- Electrical Circuit
VALUE CREATED
ThirdEye has developed a complete system, running on the production floor as per business requirements. The AI-powered system is delivering results with 85% of accuracy to detect plywood sheets with faulty measurements. The customer is happy with the result and asked ThirdEye to work further on the project to improve the accuracy rate.
Transforming Enterprises with
Data & AI Services & Solutions.
ThirdEye delivers Data and AI services & solutions for enterprises worldwide by
leveraging state-of-the-art Data & AI technologies.
