Machine Vision Based Flange Dimension Detection System for Automotive Pipe Fittings

Literature Overview

This paper by Zhao Fengsheng, Yuan Haibing, Wu Jun, and Xia Lei, published in Electronic Measurement Technology (2023, Vol. 46, No. 2, pp. 154-160), presents the development and validation of a machine vision-based dimension detection system for automotive pipe fitting flanges. The research is supported by the Ministry of Education Industry-University Cooperation Collaborative Education Project (201902016046, 201902118034) and the Shiyan Science and Technology Research Guidance Project (16Y97).

Core Technical Content

The study addresses a significant quality control challenge in automotive pipe fitting manufacturing: the inefficiency and high false detection rate associated with traditional manual sampling inspection of flange dimensions. The developed system employs machine vision technology to achieve automated, high-precision dimensional measurement of flange end-face features.

System Architecture and Processing Pipeline

The detection system follows a structured image processing workflow:

1. Camera Calibration: Performed using HALCON software to establish the spatial relationship between image coordinates and real-world dimensions.
2. Image Preprocessing: Median filtering is applied for noise removal, followed by grayscale conversion for image enhancement.
3. Edge Detection: The Canny operator is employed for sub-pixel edge detection, providing precise localization of flange boundaries.
4. Contour Fitting: Least squares fitting is applied to the extracted sub-pixel edge information to generate continuous hole diameter contours.
5. Measurement and Display: HALCON and C# joint programming is used to display measurement results in real-time.

Performance Metrics

Quality Control Integration and Engineering Practice

From a quality control perspective, this system represents a significant advancement in dimensional inspection capabilities for automotive pipe fittings. The automotive industry demands extremely tight dimensional tolerances on pipe fitting flanges to ensure proper sealing, leak prevention, and compatibility with downstream assembly processes. The 0.08 mm measurement accuracy achieved by this system is well within the typical tolerance requirements for automotive applications, which generally range from 0.1 mm to 0.5 mm depending on the specific fitting type and application.

The system's 1.2-second detection time per flange enables near-100% inspection coverage rather than the sampling inspection typically performed manually. This shift from sampling to full inspection represents a fundamental improvement in quality assurance strategy, as it eliminates the statistical uncertainty inherent in sampling methods and provides complete traceability of dimensional data for every produced part.

Application to Pipe Fitting Manufacturing

The flange dimensions detected by this system are critical quality characteristics that directly impact:

• Sealing Performance: Flange bore diameter and concentricity affect gasket seating and seal integrity
• Assembly Compatibility: Flange thickness and face flatness influence bolt-up torque and joint stiffness
• Weld Quality: Flange geometry affects the welding process parameters and weld bead formation
• Flow Characteristics: Internal bore dimensions impact fluid flow resistance and pressure drop

FMEA Considerations

Applying Failure Mode and Effects Analysis (FMEA) to the inspection process itself reveals several critical failure modes that the automated system addresses:

Technical Methodology Assessment

The choice of the Canny operator for sub-pixel edge detection is well-suited to this application. The Canny algorithm provides good edge localization with effective noise suppression, which is essential for measuring features on metallic surfaces that may exhibit surface reflections, machining marks, and other visual noise. The subsequent least squares fitting provides a robust method for extracting continuous contours from discrete edge points, effectively smoothing out minor measurement variations while preserving the true geometric shape.

The HALCON software platform is a well-established industrial machine vision solution that provides comprehensive image processing functions, camera calibration tools, and measurement algorithms. Its integration with C# programming enables flexible system architecture and user interface development, making the system adaptable to different production line configurations and operator interfaces.

Study Insights and Implications

This research demonstrates the practical viability of machine vision technology for dimensional inspection of automotive pipe fitting flanges. The achieved measurement accuracy of 0.08 mm and detection speed of 1.2 seconds per part represent a compelling improvement over traditional manual inspection methods, offering both higher accuracy and significantly greater throughput.

For engineers involved in automotive pipe fitting manufacturing, this technology should be considered as a viable option for upgrading quality control capabilities. The investment in machine vision inspection systems can be justified through reduced scrap rates, lower rework costs, improved customer satisfaction, and enhanced product traceability. The system also generates digital quality data that can be integrated into manufacturing execution systems (MES) and statistical process control (SPC) frameworks for continuous process improvement.

In conclusion, this study provides a well-documented example of machine vision technology applied to a specific industrial inspection challenge. The developed system achieves the required measurement accuracy and inspection speed for automotive production environments, representing a practical solution to a longstanding quality control limitation. Future development should focus on extending the system's capabilities to include additional quality characteristics such as surface finish, geometric tolerances, and potentially surface defect detection.