Predictive Modelling of the Effects of Processing Parameters on the Mechanical Properties of Fiber Reinforced Composites using Liquid Composite Molding
Bibliographic record
Abstract
This research investigates the efficacy of an automated Wet Compression Molding (WCM) process integrated with a 6-axis Asea Brown Boveri (ABB) robot [1] to manufacture carbon-fiber plaques. The study spans several facets including automation's impact on manufacturing efficiency and quality, the statistical significance of various processing parameters, and the predictive capabilities of machine learning models to reduce the reliance on experimental testing. This comprehensive analysis includes the setup and trials of the automated system, statistical evaluation of processing parameters, and the development and validation of predictive models. The automated WCM system, unique to Ontario, included a 6-axis ABB robot, a Long Fiber Thermoplastic (LFT) conveyor, a High-Pressure Resin Transfer Molding (HP-RTM) resin fill system, and a Dieffenbacher press equipped with a Laval mold. The automation aimed to reduce variations due to manual processing and machine errors, achieving a cycle time of less than one minute. Initial trials resulted in a 24% failure rate due to press faults and preform misalignment. Subsequent trials by other institutions demonstrated reduced failure rates, indicating successful automation. Mechanical testing verified that the plaques met target flexural modulus, flexural strength and when compared to the MatWeb database benchmark for PX35-UD300 with epoxy resin. Statistical analysis, including Analysis of Variance (ANOVA) and General Linear Model (GLM), identified significant processing parameters influencing mechanical performance. ANOVA revealed that press force and gap closure speed were not significant, while resin temperature, mold temperature, and mold curing time were influential, with respective p-values of 0.000, 0.000, and 0.032. The GLM analysis highlighted the plaques demolded with minimal force had the highest flexural strength. Thermal imaging showed that mold temperature significantly affected mechanical properties, suggesting the need for improved thermal regulation during manufacturing. Seven machine learning models were developed to predict mechanical properties based on processing parameters. Models included Multiple Linear Regression (MLR), Support Vector Regression (SVR), Gradient Boosting (GB), Random Forest Regression (RFR), and various neural network architectures. The RFR model, enhanced with hyperparameter tuning, achieved the highest predictive accuracy, explaining 45% of the variability in the data. The Functional API (FAPI) model using Keras® exhibited superior performance, with test predictions of 78.20% for flexural strength and 73.8% for flexural modulus. The integration of data from multiple institutions improved the models' predictive capabilities, demonstrating potential to reduce resource-intensive experimental testing. The automated WCM process, coupled with a 6-axis ABB robot, successfully produced high-quality plaques suitable for production trials. Statistical analysis provided insights into significant processing parameters, and machine learning models demonstrated the ability to predict mechanical properties accurately, reducing the need for extensive physical testing. Future improvements include optimizing preform clamping fixtures, resin fill wand placement, and thermal regulation during manufacturing. Enhancing machine learning models through advanced hyperparameter tuning and expanding datasets will further improve predictive accuracy and reliability. The findings from this research offer valuable insights for both industry and academia. Automation in WCM manufacturing can significantly enhance production efficiency and quality. Statistical and machine learning analyses provide robust tools for optimizing processing parameters and predicting mechanical properties, potentially transforming experimental testing approaches in advanced material development. Future research should focus on integrating additional data sources and refining predictive models to support evolving industry requirements and further reduce experimental dependencies.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".