Research on online detection of defective beer bottles

1. Basic Situation of Online Detection

Currently, beer production companies are increasingly emphasizing product quality, and more companies are using various automatic detection equipment on filling production lines to improve the reliability of inspections. The traditional method of checking empty bottles on the filling line is manual light inspection. However, on modern high-speed filling lines, due to visual fatigue and other reasons, relying solely on traditional manual light detection cannot guarantee reliability. Additionally, two important parts of the beer bottle—the bottle neck and the bottle bottom—cannot be inspected, leading to defective bottles (e.g., bottles with leaks or contaminants at the bottom) entering the market, which damages the company’s image.

2. Online Detection Basic Requirements

At present, foreign countries have relatively mature online detection systems based on image processing technology, with many products already on the market, such as Germany’s MIHO system, HEUFT system, and the OMNIVISIONI system from the US-based INDUSTRIAL DYNAMICS COMPANY, LTD.

Currently, most beer producers in China are developing production lines that exceed 40,000 bottles per hour. At such speeds, relying on manual inspection becomes extremely difficult, creating a pressing need for online detection equipment for empty bottles. However, foreign products are expensive and difficult to maintain, making them prohibitively expensive for most small and medium-sized breweries.

3. Objectives of Online Detection

Based on the analysis of foreign machine models and the needs and technological capabilities of small and medium-sized enterprises in China, the following objectives for online detection have been determined:

3.1 Online Detection Modules:

The basic modules include bottle neck and bottom detection. Additional modules, such as inner and outer wall detection, can be reserved for future use.

3.2 Online Detection Speed:

For bottles with a neck diameter of 64mm, the speed requirement is 42,000 bottles per hour.

3.3 Bottle Bottom Detection Resolution and Rejection Rate:

For non-marked and non-worn bottles:

When defect size is 2.5mm × 2.5mm, the rejection rate is 98%.

When defect size is 3.0mm × 3.0mm, the rejection rate is 99%.

For marked, worn, dark-colored, or color-differentiated bottles:

When defect size is 3.5mm × 3.5mm, the rejection rate is 98%.

When defect size is 4.0mm × 4.0mm, the rejection rate is 99%.

The system can process mixed-color or dark bottles.

4. Online Detection Solution

4.1 Proposed Solutions:

Solution 1: Directly purchase foreign machine vision systems and develop custom software on top of them. This solution has a fast development speed, but machine vision systems are expensive, and software registration fees must be paid for each system produced. Also, few existing systems can meet the speed requirements, making this solution impractical.

Solution 2: Adopt a hierarchical control system, with an industrial control computer for upper-level control and management, and a DSP chip-based system for image acquisition and processing. Although the development cycle is long, and this solution is based on fully mature algorithms, it is more suitable as a future direction, but is postponed for now.

Solution 3: Use an image acquisition card and industrial control computers with self-developed control software. Multiple systems are used to meet processing speed requirements. This solution is economically, technically, and practically more suitable for small and medium-sized enterprises. Some breweries have successfully implemented this approach, making it the chosen solution.

4.2 Online Detection Workflow:

The empty bottles to be tested are sent by a conveyor belt. When they reach the mixed bottle and color difference compensation light source, the light flashes, and the mixed bottle and color difference sensor detect transmitted light. The results are sent to the color difference compensation circuit, which adjusts the brightness of the light source for bottom detection. When the empty bottle reaches the bottom detection position, the light flashes, and the shutter opens. The imaging system captures the image, which is sent via the image acquisition card to the bottom computer for processing. If a defect is detected, a rejection signal is sent through a 32-channel I/O card. This signal is amplified and used to control the rejection mechanism, pushing the defective bottle off the conveyor. The same process is applied for bottle neck detection.

The main control computers for bottom and neck detection communicate via serial ports to ensure system coordination. Both systems use the same touch screen for display and operation, and the switching circuit completes the switching between the two systems.

4.3 Bottle Neck and Bottom Processing Module Working Principle:

4.3.1 Bottle Neck Detection:

A special imaging device is used to illuminate the bottle neck’s sealing surface with a flash. A perfect, undamaged bottle neck will reflect light as a smooth and complete ring. If the ring is broken (through cracks), or the inner edge is concave (damage on the inner edge), or the outer edge is convex (damage on the outer edge), it will be detected. Users can define three threshold values for the severity of breakage, concave, or convex deformation. The system compares the actual defect size with the preset threshold to decide whether to reject the bottle. In order to avoid misjudgment caused by the bottle's tilt, the module also includes an automatic tracking function for the bottle neck position.

4.3.2 Bottle Bottom Detection:

First, the grayscale image acquired by the image capture card is binarized. Statistical methods are then applied to calculate the size of independent contaminants. These sizes are compared with user-defined acceptable sizes to decide whether to reject the bottle.

5. Conclusion

By using image acquisition cards and multi-system industrial control computers, the issue of detecting defective beer bottles on the production line has been effectively solved at a relatively low cost, making it suitable for most small and medium-sized breweries. This solution also offers good economic benefits.