Detection of 80 Types of Capsule Defects: Enable Operators to Self - Train without Coding and Replace Manual Visual Inspection

In the pharmaceutical manufacturing industry, the accurate detection of capsule appearance defects is an important part of ensuring drug quality. However, traditional detection methods face problems such as low efficiency and poor accuracy. WeLinkirt's DaoAI AI-AOI system provides an effective solution to this problem, realizing the automation and intelligence of capsule appearance detection.

In the pharmaceutical industry, oral preparations are a common form of drugs, and capsules are particularly widely used. Capsules can not only mask the unpleasant odor of drugs but also protect drugs from the destruction of gastric acid, improving the stability and bioavailability of drugs. However, in the production process of capsules, due to various factors such as raw materials and production processes, various appearance defects may occur. For oral preparation factories, ensuring the appearance quality of capsules is a key link to guarantee the quality and safety of drugs. Traditionally, the detection of capsule appearance defects mainly relies on manual visual inspection. However, with the expansion of production scale and the improvement of product quality requirements, the limitations of manual visual inspection are increasingly prominent.

Pain Points: Why Is It Difficult?

From a quantitative perspective, manual visual inspection faces multiple challenges. First of all, there are a large number of types of capsule appearance defects. An oral preparation factory has sorted out up to 80 types of defects, including shell collapse, dents, bubbles, deformation, powder leakage, double caps, uneven lengths, color differences, and printing defects. With such a large number of defect types, operators must highly concentrate during the inspection process. However, the resolution ability and concentration time of the human eye are limited. Statistics show that after continuous work for 2-3 hours, the miss-detection rate of operators will increase significantly, reaching 10% -15%, and the false-detection rate will also increase to 5% -8%.

Secondly, it is difficult to unify the detection standards. Different operators may have different judgment standards for defects. Even the same operator may have inconsistent judgment standards at different time periods and fatigue states. This leads to the lack of reliability and consistency in the detection results. In addition, manual visual inspection cannot leave traceable judgment records. Once a quality problem occurs, it is difficult to conduct effective traceability and troubleshooting.

The root cause of the difficulty in manual visual inspection lies in the physiological and cognitive limitations of human beings. The human eye is prone to fatigue during long-time visual work, and it is difficult to accurately identify minor defects and color differences. At the same time, the speed of manual detection is relatively slow, which cannot meet the needs of large-scale production. Although traditional customized vision projects can improve the detection efficiency to a certain extent, they have a long development cycle, usually taking 3-6 months, high costs, and are difficult to adjust flexibly according to the frequent model changes of products.

Technical Principle

WeLinkirt's DaoAI AI-AOI system adopts advanced artificial intelligence algorithms and imaging technology. In terms of algorithms, the system uses deep-learning algorithms to learn and train a large number of capsule images, and can accurately identify various types of defects. Deep - learning algorithms have powerful feature-extraction and classification capabilities. They can automatically extract defect-related features from images and classify and judge defects based on these features.

In terms of imaging, the system is equipped with high-resolution industrial cameras and professional lighting equipment, which can clearly capture the appearance images of capsules. High - resolution images can provide more detailed information, which helps the algorithm to identify defects more accurately. At the same time, professional lighting equipment can eliminate interference factors such as shadows and reflections, improving the quality of images. Compared with traditional detection methods, the DaoAI AI-AOI system has higher accuracy and efficiency. Traditional methods often rely on fixed rules and templates, and it is difficult to effectively identify complex defects and new defect types. However, the DaoAI AI-AOI system can adapt to various defect situations by continuously learning and iterating the model.

Typical Application Scenarios

• Shell collapse detection: Shell collapse is one of the common defects of capsules. During the detection process, the system judges whether there is a shell-collapse phenomenon by analyzing the contour and shape of the capsule. The difficulty lies in accurately distinguishing normal slight deformation from real shell collapse, which requires the system to have high sensitivity and accuracy. The system analyzes the capsule image from multiple angles, extracts contour features, and compares them with the contour of a normal capsule. Once an obvious difference is found, it is determined as a shell-collapse defect.
• Dent detection: Dent defects are usually very subtle and difficult to detect. The system uses high-precision imaging technology and image-processing algorithms to detect subtle changes on the capsule surface. By analyzing the gray-scale value and gradient information of the image, it can find the difference between the dented part and the normal part. The difficulty lies in eliminating interference factors such as surface texture to ensure accurate detection of dent defects.
• Powder leakage detection: Powder leakage can affect the quality and safety of capsules. The system judges whether there is powder leakage by detecting powder traces on the capsule surface and color changes in the image. The difficulty lies in distinguishing normal powder residues from real powder-leakage situations, which requires comprehensive judgment based on the capsule production process and image features.
• Printing defect detection: Printing defects such as unclear printing and printing errors can affect the identification and instructions of drugs. The system uses character-recognition and image-matching technology to detect the clarity, integrity, and accuracy of printing. The difficulty lies in dealing with different fonts, font sizes, and colors of printing, as well as coping with problems such as blurred printing and reflections.

Implementation Case

A medium-sized oral preparation factory with a daily capsule production of about 500,000 capsules and a wide variety of products has high requirements for the accuracy and efficiency of capsule appearance detection. Before introducing WeLinkirt's DaoAI AI-AOI system, the factory mainly relied on manual visual inspection, with low detection efficiency and high miss-detection and false-detection rates. During the implementation process, WeLinkirt's technical team provided systematic training to the operators, enabling them to proficiently use the zero-code self-training function. Operators can mark samples, train, and iterate the model by themselves through the graphical interface, and completed the establishment of the 80-type defect detection model in a short time.

Truly sustainable automation is to return the ability of training and model change to the operators who know the products best.

Before the implementation, the miss-detection rate of manual visual inspection was about 12%, the false-detection rate was about 7%, and the detection speed was 3000-4000 capsules per hour. After the implementation, the miss-detection rate of the DaoAI AI-AOI system was reduced to less than 1%, the false-detection rate was reduced to less than 2%, and the detection speed was increased to more than 10,000 capsules per hour. At the same time, the model-change time was shortened from several hours to 5 minutes, greatly improving the production efficiency.

WeLinkirt's Solution and Product

The core advantage of WeLinkirt's DaoAI AI-AOI system lies in zero-code self-training. Operators on the production line can mark samples, train, and iterate the model by themselves through the graphical interface without programming or calling engineers. This enables operators to adjust the detection model in a timely manner according to the actual production situation and quickly adapt to new defect types and product specifications. With the 5-minute zero-code model-change function, when the product specifications are switched, operators can complete the model-change operation on - site without stopping the production line for development, greatly improving the flexibility and efficiency of production.

This system can cover all 80 types of capsule defects in a single system, realizing the full automation of capsule appearance detection. At the same time, the system keeps records throughout the judgment process, and all detection results are detailed, facilitating quality traceability and data analysis.

Quantitative Results

After using the DaoAI AI-AOI system, the capsule appearance detection in the factory has been upgraded from manual visual inspection to AI-AOI full inspection, achieving remarkable quantitative results. The miss-detection rate has been reduced from 12% to less than 1%, a decrease of more than 90%; the false-detection rate has been reduced from 7% to less than 2%, a decrease of about 70%. The detection consistency has been greatly improved, and the production capacity has been increased from 3000-4000 capsules per hour to more than 10,000 capsules per hour, a 2-3-fold increase. Operators have shifted from repetitive visual inspection work to model maintenance and abnormal handling, improving both work efficiency and quality.

This article was generated by AI. Customer cases are simulated scenarios based on real product capabilities and figures are illustrative; see product pages for official benchmarks.