TL;DR
This paper tackles the challenge of controlling radial runout in remanufactured crankshafts, a critical quality metric affected by the inherent uncertainties in recycled parts and remanufacturing processes. The authors propose a novel non-linear model that quantifies the relationship between the uncertainty in crankshaft journal roundness and bending, and the resulting radial runout acceptance rate, offering a data-driven approach to optimize remanufacturing quality.
What Problem Does This Paper Address?
The remanufacturing of crankshafts, while environmentally and economically beneficial, faces significant quality control hurdles. A key issue is achieving acceptable radial runout, a measure of how much the crankshaft deviates from a perfect axis of rotation. This problem is exacerbated by two main factors:
- Uncertainty in Recycled Parts: Crankshafts are recovered after varying service lives and conditions, leading to diverse levels of wear, deformation, and potential damage. This makes the initial quality of the "raw material" for remanufacturing highly uncertain.
- Variability in Remanufacturing Processes: Different remanufacturing routes, processes, and methods are employed depending on the condition of the recycled crankshaft. This inherent variability introduces further uncertainty into the manufacturing outcome.
Traditional remanufacturing often aims to meet new part standards, but this can lead to low acceptance rates for radial runout. The paper identifies that the roundness of main journals and the bending of the crankshaft are the primary factors influencing radial runout. However, quantifying and managing the uncertainty associated with these factors and their impact on the final radial runout quality is a complex challenge.
Methods
The authors introduce a multi-step methodology to address the problem:
- Uncertainty Measurement Model: They first define and quantify the uncertainty associated with two key parameters:
- Main Journal Roundness: This is measured by comparing the actual roundness of each journal to its ideal value (zero deviation). The uncertainty is quantified using a discrete probability distribution based on measurement tool precision and the range of observed roundness values.
- Crankshaft Bending: Similar to roundness, bending uncertainty is measured by comparing observed bending to the ideal zero value. This is also quantified using a probability distribution based on measurement precision and observed bending ranges.
-
Non-linear Reflection Relationship Model: A core contribution is the development of a non-linear model that maps the uncertainties in journal roundness and crankshaft bending to the acceptance rate of radial runout. They propose a cubic non-linear model of the form: $Y = \beta_0 + \beta_1\phi + \beta_2\phi^2 + \beta_3\phi^3$ where $Y$ is the radial runout acceptance rate, and $\phi$ represents the combined uncertainties from journal roundness and bending.
-
Gauss-Newton Iteration: To solve this non-linear model and determine the fitting coefficients ($\beta$ values), the authors employ the Gauss-Newton iterative method. This is a standard technique for non-linear least squares problems, chosen for its relative simplicity and efficiency with appropriate initial values.
- Influence Coefficient Calculation: The paper also details a method to calculate "influence coefficients" for each uncertainty factor. This allows them to analyze how much each specific uncertainty (e.g., roundness of a particular journal, or overall bending) contributes to the variation in radial runout acceptance rate. This is crucial for identifying which factors to prioritize for control.
Datasets
The study utilizes data from 20 batches of remanufactured crankshafts produced by a specific remanufacturing company. This dataset includes measurements of:
- Main journal roundness (in µm)
- Crankshaft bending (in µm)
- Radial runout acceptance rate (in %)
The paper notes that the measurements were taken using a coordinate measuring machine (0.9 µm precision) and a measurement platform with a dial indicator (1 µm precision).
Key Insights and Ideas
- Quantifying Uncertainty is Key: The paper demonstrates that simply measuring parameters isn't enough; quantifying the uncertainty around those measurements is critical for understanding their impact on final product quality.
- Non-linear Relationships Matter: The relationship between uncertainty factors and radial runout acceptance is not linear. A cubic model effectively captures this complex interaction, providing a more accurate predictive tool.
- Prioritization for Improvement: The influence coefficient analysis is a powerful tool. It allows remanufacturers to identify which specific uncertainties (e.g., bending, or roundness of certain journals) have the most significant negative impact on radial runout acceptance. This enables targeted process improvements and resource allocation.
- Balancing Quality and Cost: By understanding which uncertainties have less impact, remanufacturers can relax control requirements for those factors, potentially reducing costs and resource consumption without significantly compromising overall quality.
- Data-Driven Process Optimization: The developed model provides a framework for continuous improvement. By regularly updating the model with new data, remanufacturers can adapt to process changes and optimize their operations over time.
Key Results
- The proposed non-linear model, solved using Gauss-Newton iteration, demonstrated good fitting accuracy with small residual errors (as shown in Table 3).
- Analysis of influence coefficients (Table 4) revealed that crankshaft bending uncertainty had the largest negative impact on radial runout acceptance rate, followed by the roundness uncertainty of main journals ③, ②, and ④. Journals ① and ⑤ had a smaller influence.
- Based on these findings, the authors helped the remanufacturing company establish new technical standards:
- Bending uncertainty: Mean controlled within 5 µm, max 6 µm.
- Journals ③, ②, ④ roundness uncertainty: Mean controlled within 5 µm, max 5.5 µm.
- Journals ①, ⑤ roundness uncertainty: Relaxed requirements for parts with initial roundness below 5 µm, with repaired parts aiming for a mean of 5 µm and a max of 6.5 µm.
- Implementation of these new standards led to a significant increase in radial runout acceptance rate from an average of 84.22% to 88.35%, a 4.13% improvement.
- This quality improvement also resulted in a reduction in after-sales warranty claims due to remanufactured crankshaft defects by 697,000 RMB over six batches, demonstrating substantial economic benefits.
Limitations and Future Work
- Model Specificity: The model was developed and validated using data from a single remanufacturing company. Its generalizability to other companies with different processes, equipment, or recycled part sources may require further investigation.
- Data Requirements: The accuracy of the model relies on sufficient and representative data. Collecting and processing this data can be resource-intensive.
- Dynamic Processes: The paper acknowledges that processing technologies can change over time. It suggests that the model coefficients need to be continuously updated with new measurement data to maintain accuracy and reflect evolving production realities.
- Other Uncertainty Factors: While the paper focuses on journal roundness and bending, other factors might also influence radial runout. Future work could explore incorporating these additional uncertainties into the model.
- Broader Application: The methodology could potentially be extended to other remanufactured components where similar uncertainty-driven quality control challenges exist.
Summary updated