In the field of advanced manufacturing, particularly for aerospace components, the lost wax investment casting process is critical for producing high-integrity parts such as turbine blades and guides. However, traditional methods rely heavily on empirical trial-and-error approaches, leading to prolonged optimization cycles, high labor costs, and inefficiencies. To address these challenges, we have developed an integrated computational platform that leverages finite element simulation, design of experiments (DOE), and multi-objective optimization algorithms. This platform automates the entire process from parameter design to result analysis, significantly enhancing the efficiency and accuracy of optimizing lost wax investment casting processes. In this article, we detail the platform’s architecture, functionality, and application through a case study on a turbine guide component, demonstrating its capability to reduce defects like shrinkage porosity while minimizing computational time.
The core of our platform is built on the principles of Integrated Computational Materials Engineering (ICME) and Materials Informatics, which emphasize data-driven approaches and high-throughput computing. By integrating ProCAST finite element simulation software with Python-based automation scripts, we enable seamless execution of multiple simulation tasks in parallel. The platform supports various DOE algorithms, such as Latin Hypercube, Box-Behnken, and Central-Composite designs, to systematically explore the parameter space. Additionally, optimization algorithms like NSGA-II, NLPQLP, and Multi-Island Genetic Algorithm are incorporated to identify optimal工艺 parameters. This holistic approach allows for rapid iteration and validation, moving beyond the limitations of manual methods in lost wax investment casting.
The system architecture of our platform is divided into three layers: presentation, business, and data layers. The presentation layer utilizes HTML5, CSS, and JavaScript for user interface rendering, ensuring cross-browser compatibility. The business layer handles key functionalities, including template management, DOE-based experimental design, job scheduling and monitoring, result post-processing, and optimal solution screening. The data layer provides support through database operations, caching, and stored procedures. This layered design facilitates scalability and ease of use, allowing multiple users to submit and monitor jobs simultaneously. Moreover, the platform operates on both Windows Server and Linux systems, accommodating diverse computational needs. The integration of these components ensures that the platform can handle complex simulations typical in lost wax investment casting, such as those involving heat transfer, fluid flow, and solidification phenomena.
To illustrate the platform’s functionality, we describe its three main modules: Integrated Computation, Integrated Results, and Job Management. The Integrated Computation module allows users to select process variables, such as pouring temperature, ambient temperature, and thermal emissivity, and automatically generates试验 designs using DOE algorithms. For instance, in lost wax investment casting, these parameters critically influence defect formation. The module then creates ProCAST simulation scripts and initiates parallel computations on high-performance servers. The Integrated Results module processes the simulation outputs, extracting key metrics like shrinkage volume from STL files and constructing surrogate models. These models, often expressed as response surfaces, enable quick predictions and optimizations. The Job Management module oversees task scheduling and real-time progress monitoring, providing visual feedback to users. This modular design streamlines the entire workflow, from initial setup to final optimization, reducing human intervention and accelerating discovery in lost wax investment casting processes.
We applied the platform to optimize the lost wax investment casting of a turbine guide component, made from K418B superalloy, which is commonly used in aerospace engines. The component features a complex geometry with 58 blades and thin sections, making it prone to shrinkage defects. Our goal was to minimize shrinkage volume by adjusting three key parameters: pouring temperature, ambient temperature, and thermal emissivity. The ranges for these variables were set based on industrial practices, as shown in Table 1.
| Variable | Lower Bound | Upper Bound |
|---|---|---|
| Pouring Temperature (°C) | 1450 | 1550 |
| Thermal Emissivity | 0.5 | 0.7 |
| Ambient Temperature (°C) | 50 | 100 |
Using the Latin Hypercube DOE algorithm, we generated 24 experimental designs to explore the parameter space efficiently. The platform automated the simulation process, including mesh generation, boundary condition setting, and solver execution in ProCAST. Each simulation modeled the solidification process, and the results were automatically extracted to compute shrinkage volumes. The data from these simulations are summarized in Table 2, which includes the input parameters and corresponding shrinkage volumes. This dataset served as the basis for building an approximate model to predict shrinkage behavior in lost wax investment casting.
| Experiment No. | Pouring Temperature (°C) | Ambient Temperature (°C) | Thermal Emissivity | Shrinkage Volume (cm³) |
|---|---|---|---|---|
| 1 | 1498 | 54.35 | 0.6304 | 16.847 |
| 2 | 1528 | 95.65 | 0.5957 | 20.950 |
| 3 | 1454 | 76.09 | 0.5435 | 16.358 |
| 4 | 1502 | 67.39 | 0.6913 | 22.283 |
| 5 | 1550 | 60.87 | 0.5783 | 15.887 |
| 6 | 1459 | 56.52 | 0.6043 | 15.362 |
| 7 | 1511 | 69.57 | 0.5870 | 18.504 |
| 8 | 1463 | 63.04 | 0.6826 | 20.291 |
| 9 | 1537 | 71.74 | 0.6478 | 15.514 |
| 10 | 1502 | 84.78 | 0.6565 | 23.734 |
| 11 | 1480 | 58.70 | 0.5348 | 19.165 |
| 12 | 1485 | 100.00 | 0.6391 | 23.497 |
| 13 | 1546 | 80.43 | 0.5522 | 17.796 |
| 14 | 1476 | 73.91 | 0.6217 | 20.755 |
| 15 | 1524 | 65.22 | 0.5087 | 17.126 |
| 16 | 1541 | 91.30 | 0.6739 | 17.592 |
| 17 | 1515 | 50.00 | 0.5609 | 19.347 |
| 18 | 1450 | 89.13 | 0.6043 | 16.355 |
| 19 | 1467 | 97.83 | 0.5261 | 19.800 |
| 20 | 1489 | 76.09 | 0.5000 | 20.189 |
| 21 | 1520 | 93.48 | 0.5174 | 21.299 |
| 22 | 1493 | 86.96 | 0.5696 | 19.084 |
| 23 | 1472 | 82.61 | 0.7000 | 21.249 |
| 24 | 1533 | 52.17 | 0.6652 | 15.469 |
From the simulation data, we derived a surrogate model using regression analysis to represent the relationship between the process variables and shrinkage volume. The model is given by the equation:
$$ Y = 20.02 – 0.4879x_1 + 2.07x_2 + 0.7023x_3 + 0.1656x_1x_2 – 1.92x_1x_3 + 0.5461x_2x_3 – 4.44x_1^2 + 0.476x_2^2 + 10.2x_3^2 $$
where \( Y \) is the shrinkage volume in cm³, \( x_1 \) is the pouring temperature in °C, \( x_2 \) is the ambient temperature in °C, and \( x_3 \) is the thermal emissivity. This quadratic model had a coefficient of determination (R²) of 0.7693 and a p-value of 0.0032, indicating statistical significance and good predictive capability for lost wax investment casting optimization. The model captures nonlinear interactions among the variables, which are common in thermal processes like solidification.
We then employed the Multi-Island Genetic Algorithm (Multi-Island GA) to optimize the surrogate model, aiming to minimize shrinkage volume. After 1000 iterations, the algorithm converged to an optimal solution: pouring temperature of 1549.5°C, ambient temperature of 50.45°C, and thermal emissivity of 0.686. At these settings, the predicted shrinkage volume was 12.721 cm³, which was validated through an additional simulation. This represents a significant improvement over initial trials, highlighting the platform’s effectiveness in identifying optimal parameters for lost wax investment casting.
The efficiency gains from using our platform are substantial. Traditionally, conducting 24 simulation trials manually would require approximately 24 hours, considering setup, execution, and analysis time. With our integrated approach, the entire process—including DOE generation, parallel simulation, result extraction, and optimization—was completed in just 2 hours, representing a 91.66% reduction in time. This acceleration is achieved through automation and parallel computing, which are essential for handling the complexities of lost wax investment casting, such as multiple iterations and large datasets. The platform’s ability to automate repetitive tasks and leverage high-performance computing resources makes it a powerful tool for industrial applications, where rapid prototyping and cost reduction are critical.
In terms of practical validation, the optimized parameters were applied in a production environment for the turbine guide component. The resulting castings underwent X-ray inspection and met the HB20040-2011 standard for aerospace-grade superalloy castings, with minimal defects. This success demonstrates the platform’s reliability in translating virtual optimizations to real-world lost wax investment casting processes. The integration of simulation and optimization not only enhances quality but also supports sustainable manufacturing by reducing material waste and energy consumption.
Looking ahead, we plan to expand the platform’s capabilities by incorporating more advanced machine learning algorithms and real-time data from sensors in casting facilities. This will enable adaptive control and further refinement of lost wax investment casting processes. Additionally, we aim to include more material systems and defect types, such as hot tearing and misruns, to broaden the platform’s applicability. The ongoing development aligns with the goals of Materials Genome Initiative, promoting data-driven innovation in materials engineering.
In conclusion, our integrated computational platform represents a significant advancement in the optimization of lost wax investment casting processes. By combining DOE, finite element simulation, surrogate modeling, and multi-objective optimization, we have created a system that reduces reliance on empirical methods and accelerates process development. The case study on the turbine guide component underscores the platform’s ability to minimize defects like shrinkage porosity while drastically cutting computational time. As industries move towards smarter manufacturing, such tools will play a pivotal role in enhancing efficiency, quality, and sustainability in lost wax investment casting and beyond.