The relentless pursuit of higher thrust, superior performance, and enhanced fuel efficiency in modern aircraft engines places unprecedented demands on the design and manufacturing of critical components. Hot-section parts such as turbine blades and casings are primarily formed via vacuum investment casting of superalloys. The intricate internal channels and extensive thin-walled structures inherent to these components make the control of microstructure, internal metallurgical defects, and dimensional distortion exceptionally challenging. These factors critically constrain the yield rate, in-service performance, and operational lifespan of aerospace castings.
Superalloy vacuum investment casting is a quintessential high-temperature thermal processing operation. Often termed precision casting due to minimal machining allowances on large surface areas, it involves processing temperatures exceeding the alloy’s melting point, typically between 1,500°C and 1,600°C. This far exceeds the temperatures used in forging. The melt mass, usually ranging from 20 to 100 kg, is significantly larger than in welding processes. The alloy systems themselves are complex, comprising over a dozen elements (Ni, Fe, Co, Cr, Al, Ti, Nb, Mo, Re, C, B, Zr, etc.), which is far more than other common structural alloys like aluminum or titanium alloys. The process chain encompasses dozens of steps—from pattern wax injection and shell building to preheating, pouring, and post-processing—involving an estimated hundred distinct process parameters. Quality issues are multifaceted and multi-scale, spanning from nanoscopic to macroscopic levels, and include micro/macro-cracks, micro/macro-porosity, localized and global deformation, as well as non-conforming microstructure (grain size, inclusions) and mechanical properties. Consequently, superalloy precision casting is a prominent example of a high-temperature, multi-physics, multi-component, multi-process, and multi-scale challenge. Compounded by the geometric complexity of aerospace castings—large thin-walled areas, spatially curved surfaces, complex hollow channels, and significant wall-thickness gradients—the production of high-integrity superalloy castings faces numerous significant hurdles.
Based on extensive research and production practice in the liquid forming of aerospace superalloy components, this article analyzes the core challenges in this field. Against the backdrop of the global manufacturing digital transformation, it explores the feasibility and specific technical pathways for digitizing superalloy precision casting. Through the design and application of digital solutions for pilot product key processes, we elucidate the critical factors behind prevalent quality issues like solidification defects, dimensional inaccuracies, and substandard microstructure/properties. The aim is to demonstrate how digital solutions can effectively control these quality problems, thereby providing a reference framework for the industrial digital transformation of aerospace superalloy precision casting.
1. Overview of Industrial Digital Transformation
The advancement of industrial technology has ignited a profound “Industrial Digital Transformation” wave across global manufacturing enterprises. From Fortune 500 companies to small and medium-sized firms, the primary goal is to develop a digital economy, promote digital enablement, and achieve digital gains, inevitably leading to productivity evolution and methodology innovation. Pioneers like General Electric (GE), Siemens, and Schneider Electric have led bold initiatives. GE introduced the Industrial Internet concept early on, combining software, hardware, and analytics into manufacturing solutions. Siemens integrated the “Industry 4.0” concept and demonstrated early applications of Cyber-Physical Systems (CPS) with technologies like robotic digital twins. Schneider Electric successfully deployed digital solutions across its supply chain and manufacturing facilities. These corporate案例 provide a valuable reference for superalloy foundries embarking on their digital journey.
As this transformation accelerates, a suite of enabling generic technologies has emerged, including the Industrial Internet of Things (IIoT), Artificial Intelligence (AI), Digital Twin (DT), 5G, and Blockchain. These are rapidly permeating various industrial segments, including final assembly, component manufacturing, and raw material processing. Among these, the combined application of IIoT and AI is the most mature and successful, primarily used for automating and digitizing workshop, warehouse, and logistical processes. For instance, IIoT+AI enables product “online monitoring” and “closed-loop control,” significantly enhancing automation levels and quality inspection efficiency. IIoT handles communication, signal acquisition, analog-to-digital conversion, data preprocessing/storage, and signal feedback between equipment and control centers. AI, combined with traditional modeling methods (e.g., physics-based models, linear models, statistical models), accomplishes data-driven modeling and analysis tasks.
The Digital Twin concept, defined as the bidirectional flow of information between a physical entity and its virtual counterpart (the digital twin model), is gaining traction. While standards are emerging, developing a digital twin suitable for specific production scenarios remains a key focus and challenge for industrial enterprises. More importantly, objectively measuring the enablement and value (digital gain) delivered by DT and other digital technologies is a prerequisite for their widespread adoption and scaling within a company.
Within this industry-wide shift, the digital transformation of the casting industry emphasizes green, environmentally friendly, and energy-saving practices, addressing its historical reputation for high pollution and resource consumption. In aerospace superalloy investment casting, significant environmental and efficiency challenges exist in pattern making, shell building, pouring, and post-processing. Most critically, casting defects leading to low yield rates result in tremendous material waste, energy loss, and high production costs. Traditional process optimization, based on experience and flow simulation, faces limitations like high cost, low efficiency, model complexity, and limited predictive accuracy. Therefore, the digital transformation of superalloy precision casting is not just beneficial but essential.
Following the fundamental logic of industrial digital transformation, developing a roadmap for superalloy casting must consider digital enablement and digital gain to ultimately achieve a digital economy. For enablement, the focus is on tackling long-standing pain points: quality control difficulties, low yield, high cost, low production efficiency, and long lead times. Technologies like IIoT, AI, and DT are well-suited for these areas. Thus, key development directions include building IIoT for critical process steps, applying digital modeling to uncover the root causes and control methods for quality issues, and implementing feedback control via digital twins. The organic integration and application of IIoT+AI+DT can help achieve digital gains in quality control, cost control, and production cycle management for superalloy investment casting products during both R&D and batch production. For example, collecting product design, production, and quality data via IIoT, followed by data-driven modeling to understand defect origins, can reduce reliance on personnel expertise and help solve the problem of open-loop quality control. Enhanced automation through digital transformation also lays the groundwork for better cost control and shorter lead times in stable batch production.
2. Digital Transformation for Superalloy Precision Casting
Given that superalloy precision casting is a high-temperature, multi-physics, multi-material, multi-process, and multi-scale endeavor, process optimization and quality control remain heavily reliant on empirical knowledge and generalized empirical formulas. Traditional modeling and optimization methods often struggle to find effective linear optimal solutions, necessitating a paradigm shift in the approach to problem-solving. The development and application of digital technologies not only facilitate the exploration of new pathways for quality control but also pave the way for the comprehensive digital transformation of the precision casting process itself.
The core challenges in producing high-quality aerospace castings can be summarized as follows:
| Challenge Category | Specific Manifestations | Scale | Root Cause Complexity |
|---|---|---|---|
| Solidification Defects | Micro/Macro-porosity, Shrinkage, Hot Tears | Micro to Macro | Thermal gradients, feeding dynamics, alloy composition |
| Dimensional Inaccuracy | Warpage, Distortion, Deviation from CAD | Macro | Wax pattern stability, shell restraint, thermal stress during cooling |
| Microstructure & Property Issues | Non-uniform Grain Size, Freckles, Incorrect Phase Formation, Low Mechanical Properties | Micro to Meso | Local solidification conditions, cooling rates, alloy segregation |
| Process Control | Parameter Interdependence, High Variability | Process-scale | Multi-step process chain with hundreds of interacting parameters |
A comparison between traditional and digital-era approaches highlights the shift required:
| Aspect | Traditional Approach | Digital-Transformation Approach |
|---|---|---|
| Knowledge Base | Empirical rules, offline simulation (e.g., ProCAST) | IIoT data streams, AI/ML models, real-time digital twins |
| Optimization Cycle | Trial-and-error, lengthy physical trials | Virtual prototyping, predictive analytics, rapid iteration |
| Quality Control | Post-process inspection (NDT), statistical sampling | In-line monitoring, predictive defect detection, closed-loop control |
| Decision Making | Expert-dependent, reactive | Data-driven, proactive and prescriptive |
2.1. Industrial IoT Architecture for Critical Aerospace Casting Processes
From practice, the key process steps most influencing final casting quality are wax injection, shell building, and pouring. Establishing an IIoT framework enables hardware networking, monitoring, and signal control for these steps.
Wax Injection Process IIoT: The system monitors injection equipment and alarms, and uses cameras to inspect pattern surface quality and process-induced damage. Thermocouples monitor mold and machine platen temperatures to prevent dimensional inaccuracies from overheating or overcooling. After injection and assembly into a complete wax cluster, coordinate measuring machines (CMM) or laser scanners connected to the network precisely measure the cluster’s dimensional accuracy.
Shell Building Process IIoT: Since robotic arms are the primary units for slurry dipping and sand stuccoing, PLCs and displacement sensors monitor their trajectory to prevent damage from improper acceleration. Cameras inspect pattern surface quality before dipping and detect shell surface cracks after stuccoing, preventing defective shells from proceeding. The fluid dynamics during slurry dipping involve complex three-phase (solid-liquid-gas) flow. While full CFD simulation is computationally heavy, the IIoT provides the real-world data stream essential for building a data-driven surrogate model of the process, which can be represented conceptually by tracking parameters like viscosity $\eta(t)$ and dip velocity $\vec{v}_{dip}$.
Pouring Process IIoT: This platform is crucial. Thermocouples monitor the shell firing and transfer cooling curves. Non-contact infrared pyrometers measure temperature profiles during shell transfer and pouring. Flow meters measure pouring speed. All data is streamed to a data hub for real-time analysis and defect prediction. Crucially, data from non-destructive testing (NDT) of final castings is fed back to build data-driven models linking process parameters to defects and distortion.
The fundamental control equation governing the thermal state during pouring and solidification, which the IIoT aims to capture in real-time, can be simplified as a heat transfer problem:
$$ \rho c_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + \dot{q}_{latent} $$
where $\rho$ is density, $c_p$ is specific heat, $T$ is temperature, $t$ is time, $k$ is thermal conductivity, and $\dot{q}_{latent}$ is the latent heat release rate due to phase change. The boundary conditions (shell temperature, heat transfer coefficient) are critical inputs provided by IIoT sensors.
2.2. Digital Twin Design for Key Process Steps
The goal of building a digital twin for superalloy investment casting is to enable detection, modeling, and control of critical processes by acquiring key parameters. This involves monitoring and visualizing key parameters, performing modeling/simulation based on design and sensor data, optimizing the component and process design via simulation, and ultimately controlling porosity and distortion. Focusing on the three key steps (wax injection, shell building, pouring) is both practical and effective.
| Process Step | Digital Twin Core Function | Key IIoT Data Inputs | Modeling Aspect |
|---|---|---|---|
| Wax Injection | Dimensional fidelity & assembly deformation control | Mold temp, injection pressure, CMM/laser scan data of wax cluster | Thermo-mechanical stress model of wax assembly under gravity: $\sigma_{wax} = f(T_{wax}, E(T_{wax}), t)$ |
| Shell Building | Shell quality consistency & defect preclusion | Robotic path data, slurry viscosity/temp, camera images for crack detection | Data-driven model linking dip parameters to shell thickness and strength; Computer Vision for defect detection. |
| Pouring & Solidification | Defect & microstructure prediction and mitigation | Shell preheat temp curve, pour temp/speed, real-time cooling curves | Real-time thermal model feeding a defect propensity model (e.g., porosity predictor function $P_{porosity} = g(G, \dot{T})$, where G is thermal gradient). |
Wax Injection DT: Large, thin-walled wax patterns are prone to dimensional inaccuracies and deformation upon assembly. The DT integrates data from the IIoT monitoring platform and networked metrology tools (CMM) to track and predict pattern deformation, creating a virtual counterpart of the physical wax cluster for analysis and virtual try-out of assembly strategies.
Shell Building DT: This DT model focuses on process control and quality inspection. It uses machine vision from camera feeds to assess pattern surface quality pre-dipping and detect shell surface defects post-stuccoing. It monitors the rigid-body motion of the robotic arm to ensure repeatability. The complex fluid dynamics are handled not by solving full Navier-Stokes equations online but by using the DT as a data integrator and surrogate model executor for efficient, real-time process guidance.
Pouring Process DT: This is the most critical and complex DT. It segments control into three stages: 1) Shell Transfer: Monitors shell temperature decay after furnace exit. If temperature falls below a threshold, the system can trigger a re-heat or recommend insulation. 2) Pre-Pour in Vacuum: Monitors that both vacuum level AND shell temperature meet criteria before authorizing the pour command, preventing defects from pouring into an overly cold mold. 3) Post-Pour Solidification: Tracks the cooldown curves of the casting and shell. This data is used to predict and, where possible, actively influence solidification conditions (e.g., by modulating cooling around the riser) to achieve target grain structures and minimize stress. The entire sequence operates under the digital system’s supervision with alarm triggers for manual intervention if needed.
2.3. Online Quality Inspection and AI-Powered Defect Analysis
Porosity remains a primary defect in superalloy castings. Establishing computer vision models and assessment methods for pore detection is key to controlling the mechanical performance of aerospace castings.
Our methodology involves a multi-scale inspection approach correlated with process data from the DT/IIoT system. At the micro-scale, metallography reveals shrinkage pores, dendrite structure, and micro-porosity. X-ray digital radiography (DR) detects larger shrinkage locations. Image segmentation techniques quantify global porosity percentage. Furthermore, advanced characterization provides high-resolution images of micro-porosity, for which we train machine learning models for automatic identification and classification.
For surface defect inspection (e.g., cracks, cold shuts), we employ Convolutional Neural Networks (CNN). A typical supervised learning approach involves training the network on labeled image datasets to perform pixel-wise segmentation or classification of defects. The training minimizes a loss function, such as a weighted binary cross-entropy loss for segmentation:
$$ \mathcal{L} = -\frac{1}{N} \sum_{i=1}^{N} [w \cdot y_i \log(\hat{y}_i) + (1 – y_i) \log(1 – \hat{y}_i)] $$
where $y_i$ is the true label (defect or not), $\hat{y}_i$ is the predicted probability, $w$ is a weight to address class imbalance (few defect pixels vs. many non-defect pixels), and $N$ is the number of pixels. This enables fast, highly sensitive, and automated surface flaw detection, replacing subjective human inspection.
Finally, mechanical testing of separately cast test bars provides property data (yield strength, UTS, elongation). The ultimate goal of the digital framework is to establish predictive links, often via AI models, between the comprehensive process parameter set $\vec{P}$ captured by IIoT, the defect state $\vec{D}$ identified by automated inspection, and the mechanical properties $\vec{M}$:
$$ \vec{M} = \mathbf{F}(\vec{P}, \vec{D}) \quad \text{or ideally,} \quad \vec{M} \approx \mathbf{G}(\vec{P}) $$
where $\mathbf{F}$ and $\mathbf{G}$ are complex, non-linear functions learned from historical production data. Optimizing $\vec{P}$ to maximize yield and $\vec{M}$ while minimizing $\vec{D}$ becomes a data-driven engineering problem.
3. Summary and Future Perspectives
Based on our practice in the liquid forming of aerospace superalloy components, we have proposed an integrated design framework combining Industrial IoT and Digital Twins. This framework aims to address core quality control challenges—such as micro-porosity and dimensional inaccuracy—in the precision casting of complex, thin-walled superalloy structures for aerospace applications, while also targeting improvements in production cost and cycle time control. The next phases of research and implementation will focus on:
1. Advanced Digital Infrastructure: Implementing 5G technology to establish a high-speed, low-latency data transmission network within the foundry for rapid transfer and storage of production data streams. This will be coupled with high-performance computing servers to execute the computationally intensive models underpinning the digital twins, enabling true real-time capability. A crucial step is the seamless integration of the digital twin system with existing enterprise resource planning (ERP) and manufacturing execution systems (MES) to close the information loop from order to delivery.
2. Deepening Defect Mechanism Understanding via AI: Developing robust AI models, potentially leveraging physics-informed neural networks (PINNs), to establish more accurate and generalizable relationships between process parameters and micro-porosity formation. This moves beyond correlation to capture underlying physical principles, enhancing the predictive power for online defect forecasting. The governing equation for pore nucleation and growth could be integrated into such a hybrid model framework.
3. High-Fidelity Distortion Prediction for Large Castings: Creating enhanced thermomechanical distortion models that explicitly account for the viscoelastic-plastic behavior of the ceramic shell and the evolving constitutive properties of the solidifying alloy. By feeding these models with highly accurate, synchronized process parameters from the IIoT (e.g., spatially resolved shell temperatures), we aim to solve the current challenge of inaccurate distortion prediction for large-scale aerospace castings, enabling proactive corrective measures in tooling or process design.
The digital transformation of aerospace superalloy casting is a continuous journey. The fusion of IIoT, Digital Twin, and AI technologies presents a transformative pathway from experience-driven art to a data-driven, predictive science. This evolution is critical for meeting the ever-increasing demands for performance, reliability, and efficiency in the next generation of aircraft engines.