Zhuojin Pipe Fitting Co., LtdMulti-Objective Prediction of Seamless Steel Tube Oblique Rolling Piercing Tube Shape
Literature Overview
This study by Jia Shiying, Wang Qinghua, Wang Zhenyan, Hu Jianhua, Shuang Yuanhua, and Zhou Xinliang from Taiyuan University of Science and Technology and Taiyuan Heavy Industry Co., Ltd. (2022), investigates the prediction of tube shape in the oblique rolling piercing process of seamless steel tubes. Supported by the Shanxi Provincial Science and Technology Major Project (20191102009), the research was published in Forging Technology, Volume 47, Issue 10, pages 169-175.
The oblique rolling piercing process is the primary method for producing seamless steel tube blanks (called "blanks" or "pierced tubes") from solid steel billets. The process involves piercing a hole through a rotating billet using a plug and two rolls arranged at an angle. The dimensions of the pierced tube, particularly the outer diameter and wall thickness, are critical for subsequent processing steps and final product quality. Accurate prediction of the tube shape is essential for process optimization and quality control.
Core Technical Content
The study addresses the challenge of predicting the pierced tube dimensions based on process parameters. The research considers production process optimization and production demand factors to develop a multi-objective prediction model based on Least Squares Support Vector Regression (LSSVR).
Factor Screening Using Grey Relational Analysis
The grey relational analysis (GRA) method was used to screen the significant process parameters that affect the tube shape. The GRA method quantifies the degree of correlation between input parameters and output variables, allowing the identification of the most influential factors. The analysis identified five key process parameters as the input variables:
| Parameter | Description | Unit |
|---|---|---|
| Front extension | Distance the plug protrudes beyond the roll gap | mm |
| Roll gap | Distance between the two rolls | mm |
| Guide plate gap | Distance between the guide plates | mm |
| Plug diameter | Diameter of the piercing plug | mm |
| Billet diameter | Diameter of the input steel billet | mm |
The output variables are the tube shape parameters:
| Parameter | Description | Unit |
|---|---|---|
| Wall thickness | Thickness of the pierced tube wall | mm |
| Outer diameter | External diameter of the pierced tube | mm |
Multi-Input Multi-Output LSSVR Model
The study constructed a multi-input multi-output (MIMO) LSSVR model to predict the tube shape. The LSSVR method is a variant of support vector regression that simplifies the optimization problem by replacing the standard support vector regression with a least squares formulation. This simplification eliminates the need for iterative optimization and provides a direct solution.
The key features of the model include:
- Multi-input capability: The model accepts five process parameters as inputs simultaneously.
- Multi-output capability: The model predicts both wall thickness and outer diameter simultaneously.
- Small sample handling: The LSSVR method is suitable for small sample sizes, which is common in industrial process data.
- Cross-correlation handling: The model accounts for the cross-correlation between input and output parameters, which is important for accurate prediction.
Model Validation
The model was validated using actual production data collected from the piercing process. The simulation experiments demonstrated the effectiveness of the model in predicting the tube shape parameters. The model provides a practical tool for process parameter adjustment and optimization in the seamless steel tube piercing production.
Interpretation of Technical Points
The oblique rolling piercing process is a complex metal forming operation that involves multiple interacting process parameters. The tube shape is determined by the interaction between the roll gap, plug diameter, front extension, guide plate gap, and billet diameter. These parameters are not independent; changing one parameter affects the others through the process mechanics. The multi-objective prediction model captures these interactions and provides a practical tool for process optimization.
The grey relational analysis is an effective method for factor screening in situations with limited data. The method quantifies the correlation between each input parameter and the output variables, allowing the identification of the most significant factors. This screening step is important for reducing the dimensionality of the problem and improving the prediction accuracy.
The LSSVR method is well-suited for the piercing process prediction problem for several reasons. First, the method can handle small sample sizes, which is common in industrial settings where data collection is limited. Second, the least squares formulation provides a direct solution without the need for iterative optimization, which is computationally efficient. Third, the method can handle non-linear relationships between inputs and outputs, which is essential for accurately modeling the complex piercing process.
The multi-input multi-output structure of the model is important because the wall thickness and outer diameter are correlated. Predicting them separately would ignore this correlation and potentially lead to inconsistent predictions. The MIMO structure ensures that the predictions are consistent with each other and with the physical constraints of the process.
Process and Standards Analysis
The prediction model should be integrated into the process control system for the oblique rolling piercing operation. The following process points should be considered:
Process Parameter Control
- Billet diameter: The billet diameter should be measured and controlled to ensure consistent input conditions. Variations in billet diameter directly affect the tube dimensions.
- Plug diameter: The plug diameter determines the inner diameter of the pierced tube. The plug should be inspected regularly for wear, as wear increases the inner diameter and reduces the wall thickness.
- Roll gap: The roll gap controls the outer diameter of the pierced tube. The roll gap should be adjusted based on the prediction model to achieve the target dimensions.
- Front extension: The front extension affects the piercing force and the tube shape. The optimal front extension depends on the billet diameter, plug diameter, and desired tube dimensions.
- Guide plate gap: The guide plate gap affects the stability of the piercing process and the tube shape. The gap should be set to provide adequate support without excessive friction.
Quality Control Integration
The prediction model can be used for real-time quality control in the following ways:
| Application | Method | Benefit |
|---|---|---|
| Process parameter setting | Model-based optimization | Reduces trial and error |
| Dimensional prediction | Real-time prediction | Early detection of deviations |
| Plug wear monitoring | Trend analysis | Predictive maintenance |
| Process anomaly detection | Deviation from predicted values | Rapid response to issues |
| Production planning | Capacity optimization | Improved efficiency |
Relevant Standards
The piercing process and the resulting tube blanks should comply with relevant standards, including:
- GB/T 8162 (Steel tubes for general cold-rolled or cold-drawn steel tubes)
- GB/T 8163 (Steel tubes for fluid transport)
- API 5CT (Specification for casing and tubing)
- ASTM A519 (Standard Specification for Seamless Carbon Steel Mechanical Tubing)
The dimensional tolerances specified in these standards should be used as the target values for the prediction model. The model should be calibrated to ensure that the predicted dimensions fall within the acceptable tolerance range.
Integration with Engineering Practice
In engineering practice, the prediction model should be used to optimize the piercing process in the following ways:
- Process parameter optimization: The model can be used to determine the optimal combination of process parameters for achieving the target tube dimensions. This reduces the need for trial and error and improves production efficiency.
- Quality control: The model can be used to predict the tube dimensions in real time, allowing early detection of deviations and rapid corrective action. This reduces the number of defective products and improves product quality.
- Plug wear management: The model can be used to monitor the effect of plug wear on the tube dimensions. When the predicted dimensions deviate from the target values due to plug wear, the plug can be replaced or the process parameters can be adjusted.
- Production planning: The model can be used to plan production runs by predicting the achievable dimensions for different billet sizes and process parameters. This improves production planning and reduces material waste.
The implementation of the prediction model in the production environment requires integration with the process control system. The model should be connected to the sensors that measure the process parameters and the actuators that control the process. The model output should be displayed on the operator interface and used to guide the process adjustment.
Key Questions and Reflections
Several questions arise from this study that warrant further investigation. First, the model was validated using a limited dataset, and its performance may vary with different production conditions. The model should be validated with a larger dataset that covers a wider range of process parameters and production conditions. Second, the model does not account for the dynamic effects of the piercing process, such as the rotational speed of the rolls and plug, the temperature of the billet, and the friction conditions. A more comprehensive model that includes these dynamic effects would provide more accurate predictions. Third, the model should be integrated with a real-time control system to enable automatic process adjustment based on the predicted dimensions.
The study also raises the question of whether the LSSVR method is the optimal choice for this prediction problem. Other data analysis methods, such as neural networks, random forests, or gradient boosting, may provide better performance. A comparative study of different prediction methods would be valuable for selecting the most suitable approach.
Study Insights and Implications
This study provides a practical tool for the optimization of the oblique rolling piercing process. The key finding is that a multi-input multi-output LSSVR model can accurately predict the tube shape parameters based on five key process parameters. The grey relational analysis provides an effective method for factor screening, and the LSSVR method is well-suited for the small sample and non-linear nature of the problem. For engineers involved in seamless steel tube manufacturing, the study underscores the importance of process parameter optimization and quality control in achieving consistent product dimensions. The prediction model can be used to reduce trial and error, improve production efficiency, and ensure compliance with dimensional tolerances specified in relevant standards.