In the aerospace industry, the demand for high-performance rocket engines with increased thrust and thrust-to-weight ratios has led to more stringent requirements for critical components, particularly casting parts such as turbine housings. These casting parts often feature complex geometries and must withstand extreme service environments. Traditional dimensional inspection methods for casting parts, including manual gauging and coordinate measuring machines (CMMs), suffer from low coverage of surface profiles, inefficiency, and high labor intensity. As a result, there is a pressing need for advanced non-contact measurement techniques to ensure precision and reliability. In this context, three-dimensional laser scanning (3DLS) technology has emerged as a powerful solution, enabling rapid, high-precision acquisition of surface data through non-contact means. This technology has found widespread applications in fields like reverse engineering, digital cities, cultural heritage preservation, and industrial metrology. In our work, we have developed an automated 3D online measurement system specifically tailored for aerospace casting parts, leveraging 3DLS to overcome the limitations of conventional approaches. This article, written from my first-person perspective as a researcher involved in the project, details the design, implementation, and testing of this system, with a focus on enhancing the inspection of casting parts.
The core challenge in inspecting aerospace casting parts lies in their intricate shapes and tight tolerances. Traditional methods often fail to provide comprehensive surface coverage, leading to potential defects going undetected. For instance, manual inspection of a casting part might only sample discrete points, leaving large areas unchecked. This not only compromises quality but also increases production costs due to rework or failures. To address this, our team embarked on designing an integrated system that combines 3D laser scanning with robotics and automation. The goal was to achieve full-surface scanning of casting parts with high accuracy and efficiency, thereby improving overall manufacturing processes. Throughout this project, we emphasized the importance of casting part integrity, as even minor deviations can impact engine performance and safety.
Our 3D online measurement system is built upon a modular architecture, comprising four main components: the measurement module, logistics module, integrated control module, and safety protection module. Each module plays a critical role in ensuring seamless operation from part identification to final reporting. Below, I will elaborate on the design principles and implementation details, incorporating tables and formulas to summarize key aspects. The system was developed with scalability in mind, allowing it to adapt to various casting part geometries common in aerospace applications.
System Design Overview
The design of our system was driven by the need for automated, high-throughput inspection of casting parts. We considered factors such as measurement accuracy, speed, and flexibility to handle different part sizes and shapes. The overall workflow begins with the arrival of a casting part at the material station, where it is identified and transported to the measurement area. A robotic arm equipped with a 3D laser scanner then performs scanning along pre-programmed paths, capturing dense point cloud data of the casting part surface. This data is processed and compared against the CAD model to generate deviation reports. To facilitate this, we integrated advanced software for data acquisition, alignment, and analysis. The design also prioritizes safety, with protective enclosures and interlocks to prevent accidents during operation. In this section, I will outline the conceptual framework and key design decisions that guided our development process.
One fundamental aspect of our design is the use of structured light 3D scanning, where a pattern of light (e.g., blue light) is projected onto the casting part surface, and cameras capture the deformations to reconstruct 3D coordinates. The accuracy of this method depends on several factors, including camera calibration, environmental conditions, and surface properties of the casting part. To quantify this, we derived a formula for the measurement error based on triangulation principles. For a point on the casting part surface, the 3D coordinates $(X, Y, Z)$ can be computed using the following equation from stereo vision:
$$ Z = \frac{f \cdot b}{d} $$
where $f$ is the focal length of the camera, $b$ is the baseline distance between the camera and projector, and $d$ is the disparity (difference in pixel locations) between corresponding points in the left and right images. The error in $Z$, denoted $\Delta Z$, can be approximated by:
$$ \Delta Z = \frac{Z^2}{f \cdot b} \cdot \Delta d $$
Here, $\Delta d$ represents the disparity error, which is influenced by factors like image noise and calibration accuracy. By minimizing $\Delta d$ through precise calibration and high-quality optics, we aimed to achieve sub-millimeter accuracy for casting part measurements. This theoretical foundation informed our selection of scanning hardware and calibration procedures.
To summarize the system design parameters, I have compiled Table 1, which lists key specifications and their target values. This table highlights how each parameter contributes to the overall performance when inspecting casting parts.
| Parameter | Target Value | Description |
|---|---|---|
| Measurement Accuracy | ± 0.05 mm | Maximum permissible error for casting part dimensions |
| Scanning Speed | 1.2 million points/second | Data acquisition rate for casting part surfaces |
| Coverage Rate | > 98% | Percentage of casting part surface area scanned |
| Part Size Range | Up to 1000 mm × 1000 mm × 1000 mm | Maximum dimensions of casting parts accommodated |
| Automation Level | Full autonomy | Degree of human intervention required for casting part inspection |
Another critical design consideration was the integration of robotics to manipulate both the scanner and the casting part. We employed two robotic arms: one for scanning (measurement robot) and one for material handling (logistics robot). This dual-robot configuration allows for optimal positioning and reduces scanning time. The measurement robot, with six degrees of freedom, can orient the scanner at various angles to capture hidden features of the casting part, such as internal passages or undercuts. Meanwhile, the logistics robot manages the loading and unloading of casting parts from material stations to a rotary table (变位机), which acts as an additional axis to rotate the part during scanning. This synergy enhances the coverage of complex casting parts without requiring manual repositioning.
In terms of software, we developed custom applications for scan planning, data processing, and analysis. The scan planning software generates robot paths based on the CAD model of the casting part, ensuring that all critical areas are covered. During data acquisition, real-time stitching algorithms merge multiple scans into a complete point cloud of the casting part. This is achieved through feature matching and iterative closest point (ICP) algorithms, which align scans using common geometric features. The analysis software then compares the point cloud to the reference CAD model, computing deviations and generating color-coded maps. For statistical process control, we also incorporated formulas to calculate key metrics like mean deviation and standard error for each casting part batch. For example, the mean deviation $\bar{\delta}$ for a set of $n$ points on a casting part surface is given by:
$$ \bar{\delta} = \frac{1}{n} \sum_{i=1}^{n} |d_i| $$
where $d_i$ is the deviation at point $i$ from the nominal surface. Such metrics help in monitoring the consistency of casting part production over time.
System Implementation and Module Details
The implementation of our system involved meticulous hardware integration and software development. I will now describe each module in detail, drawing from my hands-on experience during assembly and testing. Throughout this process, we repeatedly tested the system with various casting parts to refine its performance.
Measurement Module
The measurement module is the heart of our system, responsible for acquiring high-fidelity 3D data of casting parts. It consists of a measurement robot, a blue light 3D scanner, a calibration plate, measurement software, analysis software, and a workstation. We selected a high-precision blue light scanner due to its superior stability and accuracy for metallic surfaces common in casting parts. The scanner operates on the principle of phase-shift profilometry, where sinusoidal fringe patterns are projected onto the casting part surface, and phase information is extracted to compute depth. The calibration plate, featuring an array of known reference points, is used periodically to verify and adjust the scanner’s accuracy. During calibration, we solve for intrinsic and extrinsic camera parameters using a bundle adjustment approach, minimizing reprojection errors. This ensures that measurements of casting parts remain consistent over time.
The measurement robot, a six-axis industrial robot, is mounted on a stable base to minimize vibrations. It carries the scanner via a custom adapter, allowing precise movement along programmed trajectories. We optimized the robot paths to avoid collisions and ensure complete coverage of the casting part. The measurement software, developed in-house, provides a user interface for setting up scan projects, controlling the robot, and processing point clouds. It includes automated functions for scan alignment and filtering to remove noise from casting part surfaces. The analysis software, on the other hand, performs detailed comparisons between the scanned data and CAD models. It can generate 2D cross-sectional analyses and 3D deviation maps, outputting reports in PDF format. For instance, when inspecting a casting part like a turbine housing, the software highlights areas where thickness or contour deviations exceed tolerances.
To illustrate the components of the measurement module, Table 2 provides a summary with specifications. This table underscores how each element contributes to the precise measurement of casting parts.
| Component | Model/Specification | Role in Casting Part Inspection |
|---|---|---|
| Measurement Robot | 6-axis, payload 10 kg | Positions scanner around casting part for full coverage |
| 3D Scanner | Blue light, accuracy ± 0.02 mm | Captures surface point cloud of casting part |
| Calibration Plate | Ceramic, 300 mm × 300 mm | Ensures measurement accuracy for casting parts |
| Measurement Software | Custom C++ application | Controls scanning process and data acquisition for casting parts |
| Analysis Software | Commercial SDK integrated | Compares casting part scan to CAD model and generates reports |
| Workstation | High-performance GPU, 32 GB RAM | Processes large point clouds from casting parts |
In practice, we found that the surface finish of casting parts can affect scan quality. Rough surfaces or reflective materials may cause scattering or specular reflections, leading to data gaps. To mitigate this, we applied a thin layer of anti-reflective spray on casting parts when necessary, though this is rarely needed for matte-finished aerospace casting parts. Additionally, we implemented algorithms in the measurement software to fill small holes in the point cloud using polynomial interpolation, ensuring a complete digital model of each casting part.
Logistics Module
The logistics module automates the handling and positioning of casting parts throughout the inspection process. It includes a material handling robot, a vision positioning system, a rotary table (变位机), material stations, and transfer fixtures. The material robot, another six-axis robot, is equipped with a gripper and a 2D vision camera. It picks up casting parts from material stations and places them on the rotary table for scanning. The vision system identifies each casting part based on QR codes attached to the fixtures, enabling the system to automatically select the appropriate scanning program. This is crucial for mixed-production environments where different types of casting parts are inspected sequentially.
The rotary table serves as the seventh axis of the measurement system, allowing the casting part to be rotated during scanning. This reduces the need for the measurement robot to move extensively, thereby speeding up the process. We programmed the table to index at specific angles, ensuring that all sides of the casting part are scanned without occlusion. The material stations consist of mobile carts with custom pallets that hold casting parts in a repeatable orientation. Each pallet has locating pins and clamps to secure the casting part firmly, minimizing movement during transport. The transfer fixtures, attached to the material robot, are designed with quick-change mechanisms to accommodate different casting part geometries. We developed a library of gripper fingers that can be swapped automatically using a tool changer, enhancing flexibility for various casting parts.
To quantify the efficiency gains from automation, we derived a formula for the total inspection time $T_{\text{total}}$ of a casting part:
$$ T_{\text{total}} = T_{\text{load}} + T_{\text{scan}} + T_{\text{unload}} + T_{\text{process}} $$
where $T_{\text{load}}$ and $T_{\text{unload}}$ are the times for loading and unloading the casting part, $T_{\text{scan}}$ is the scanning time, and $T_{\text{process}}$ is the data processing time. With our automated logistics module, $T_{\text{load}}$ and $T_{\text{unload}}$ are reduced to under 30 seconds each, compared to several minutes for manual handling. This significantly boosts throughput when inspecting multiple casting parts.
Table 3 summarizes the key components of the logistics module and their functions related to casting part handling.
| Component | Function | Impact on Casting Part Inspection |
|---|---|---|
| Material Robot | Transports casting parts between stations | Reduces manual labor and handling time for casting parts |
| Vision System | Identifies casting part type and position | Enables automatic program selection for casting parts |
| Rotary Table | Rotates casting part during scanning | Improves coverage and reduces scanning time for casting parts |
| Material Stations | Holds casting parts in predefined positions | Ensures repeatable positioning of casting parts |
| Transfer Fixtures | Grips and releases casting parts | Provides secure handling for diverse casting parts |
During development, we encountered challenges in aligning the vision system for accurate part recognition. We calibrated the camera using a checkerboard pattern and implemented perspective transformation algorithms to correct for distortions. This ensured that the material robot could precisely pick up each casting part, even if placed slightly off-center on the pallet. Additionally, we designed the material stations with redundant safety sensors to detect misloaded casting parts, preventing collisions during robot motion.
Integrated Control Module
The integrated control module orchestrates the entire system, ensuring synchronization between the measurement and logistics modules. It comprises a central PLC (Programmable Logic Controller), a touchscreen HMI (Human-Machine Interface), and various I/O modules. The PLC communicates with the robots, scanner, rotary table, and sensors via industrial Ethernet protocols, enabling real-time control. We programmed the PLC to execute a sequential workflow: upon initiation, it checks the status of all subsystems, identifies the casting part via the vision system, commands the material robot to load the casting part, triggers the measurement robot to scan, and finally unloads the casting part after data processing. The touchscreen provides an intuitive interface for operators to start/stop jobs, monitor progress, and view alerts. All parameters, such as scanning speed and tolerance limits, can be configured for different casting parts through this interface.
To manage data flow, we implemented a client-server architecture where the workstation receives point cloud data from the scanner, processes it, and sends results back to the PLC for reporting. This decoupling allows for parallel processing; while one casting part is being scanned, the previous one’s data can be analyzed. We also incorporated error-handling routines to deal with common issues like scanner calibration drift or robot path obstructions. For instance, if a casting part is not detected properly, the system pauses and prompts the operator for intervention.
From a control theory perspective, we modeled the system as a discrete-event system and used state machines to manage transitions. The state of the system, denoted $S$, can be represented as:
$$ S = \{ s_1, s_2, \dots, s_n \} $$
where each state $s_i$ corresponds to a phase in the inspection cycle (e.g., idle, loading, scanning, unloading). Transitions between states are triggered by events such as sensor signals or timer expirations. This formal approach helped us design robust control logic that minimizes downtime when inspecting casting parts.
Table 4 outlines the main elements of the integrated control module and their roles in managing casting part inspections.
| Element | Specification | Role in Casting Part Inspection |
|---|---|---|
| PLC | Modular, 128 I/O points | Coordinates all actions for casting part inspection |
| HMI Touchscreen | 10-inch color display | Provides user interface for operating casting part inspections |
| Communication Network | EtherCAT protocol | Ensures fast data exchange between devices handling casting parts |
| Safety Controller | Integrated with safety relays | Monitors emergency stops during casting part handling |
During testing, we refined the control sequences to optimize cycle times. For example, we overlapped the motion of the material robot and the measurement robot where safe, reducing the overall time per casting part. The system also logs inspection data for each casting part, including timestamps and deviation statistics, which can be exported for quality assurance purposes.
Safety Protection Module
Safety is paramount in an automated system involving heavy robots and precision equipment. The safety protection module includes mechanical safety switches, emergency stop buttons, safety interlocks, and protective fencing. The entire system is enclosed within a safety cage with interlocked doors; if a door is opened during operation, the PLC immediately halts all moving components. This prevents access to hazardous areas while casting parts are being handled or scanned. We designed the safety circuits to meet PL e (Performance Level e) according to ISO 13849, ensuring high reliability. Emergency stop buttons are placed at multiple locations, allowing operators to quickly shut down the system if needed.
In addition, we installed light curtains and area scanners to detect intrusions into the robot workspace. These devices provide an additional layer of protection beyond physical barriers. For instance, if someone reaches into the cell while a casting part is on the rotary table, the system stops rotation and scanning. All safety functions are hardwired to a safety relay that directly cuts power to the robots, ensuring fail-safe behavior. We also implemented software checks, such as verifying that casting parts are properly clamped before starting the scan, to prevent accidents due to part ejection.
To quantify safety performance, we conducted risk assessments based on the formula for risk reduction factor $RRF$:
$$ RRF = \frac{1}{1 – \text{Probability of Dangerous Failure}} $$
By incorporating multiple safety layers, we achieved an $RRF$ greater than 100,000, indicating a very low probability of hazardous events during casting part inspections. This gives operators confidence when working with the system daily.
Table 5 lists the safety components and their functions in protecting both personnel and casting parts during operation.
| Safety Component | Function | Benefit for Casting Part Inspection |
|---|---|---|
| Safety Interlocks | Lock access doors during operation | Prevents unauthorized entry while casting parts are scanned |
| Emergency Stop Buttons | Immediate system halt | Allows quick response to issues with casting parts or equipment |
| Light Curtains | Detect intrusions in workspace | Protects operators during automatic handling of casting parts |
| Safety Relays | Cut power on fault detection | Ensures safe shutdown if casting part misalignment occurs |
Throughout the installation, we verified that all safety measures were functional through rigorous testing, including simulated fault conditions. This proactive approach minimized risks and ensured compliance with industrial safety standards.
System Testing and Performance Evaluation
After assembling the system, we conducted extensive tests to validate its performance on actual aerospace casting parts. The primary test object was a low-pressure oxygen pump housing casting part, a complex component with intricate internal passages and external flanges. This casting part represents typical challenges in aerospace manufacturing, requiring high dimensional accuracy. We scanned the casting part multiple times to assess repeatability and coverage.
The testing procedure began with calibrating the scanner using the calibration plate. We then placed the casting part on the material station, where the vision system identified it and triggered the automated workflow. The material robot transferred the casting part to the rotary table, and the measurement robot performed a series of scans from different angles. The rotary table rotated the casting part in 30-degree increments, ensuring full circumferential coverage. The point clouds were automatically stitched into a complete model of the casting part, which was then compared to the CAD nominal using the analysis software.
To evaluate coverage, we calculated the surface area ratio. The total surface area of the casting part, derived from the CAD model, was $A_{\text{total}} = 1,362,750.49 \, \text{mm}^2$. After scanning, the acquired point cloud covered an area of $A_{\text{scanned}} = 1,337,207.71 \, \text{mm}^2$, resulting in a coverage rate $C$ given by:
$$ C = \frac{A_{\text{scanned}}}{A_{\text{total}}} \times 100\% = 98.1\% $$
This high coverage indicates that nearly the entire surface of the casting part was captured, addressing the limitation of traditional methods that often miss areas. The missing 1.9% primarily corresponded to small, deeply recessed features that were partially occluded, but these were non-critical for dimensional checks.
For dimensional accuracy, we measured specific features on the casting part, such as diameters and heights. For example, one cylindrical section had a nominal diameter of $\phi209.5 \, \text{mm}$ with a tolerance of $\phi209_{-2}^{0} \, \text{mm}$. The measured values across multiple scans ranged from $\phi209.284 \, \text{mm}$ to $\phi209.440 \, \text{mm}$, well within tolerance. Similarly, a U-slot height with nominal $24.5 \, \text{mm}$ and tolerance $24.5_{-2}^{0} \, \text{mm}$ was measured between $24.312 \, \text{mm}$ and $24.492 \, \text{mm}$. These results demonstrate the system’s ability to verify casting part dimensions accurately.
We also assessed repeatability by scanning the same casting part ten times and computing the standard deviation of key dimensions. The repeatability error $\sigma_r$ for a diameter measurement was found to be:
$$ \sigma_r = 0.008 \, \text{mm} $$
This low value confirms that the system produces consistent results for casting parts, essential for quality control in aerospace production.
To present the test data comprehensively, Table 6 summarizes the key performance metrics for the casting part inspection.
| Metric | Value | Interpretation for Casting Part Quality |
|---|---|---|
| Surface Coverage Rate | 98.1% | High coverage ensures most of casting part surface is inspected |
| Diameter Measurement (ϕ209.5 mm) | ϕ209.284–ϕ209.440 mm | Casting part dimension within specified tolerance |
| Height Measurement (24.5 mm) | 24.312–24.492 mm | Casting part feature conforms to design requirements |
| Repeatability (Standard Deviation) | 0.008 mm | Consistent measurement across multiple scans of casting part |
| Total Inspection Time | Approx. 5 minutes | Efficient process for casting part, including scan and analysis |
In addition to quantitative metrics, we evaluated the system’s usability. Operators reported that the automated workflow reduced physical strain compared to manual inspection of casting parts. The touchscreen interface made it easy to switch between different casting part programs, and the automatic report generation saved time on documentation. We also tested the system with other casting parts, such as valve bodies and brackets, to ensure versatility. In each case, the system achieved coverage rates above 97% and maintained accuracy within ±0.05 mm, meeting aerospace standards.
The image above illustrates a typical aerospace casting part being scanned in our system, highlighting the integration of robotics and 3D scanning. Such visual documentation aids in understanding the practical application of our technology for casting part inspection.
Beyond dimensional checks, the system can detect surface defects on casting parts, such as porosity or cracks, by analyzing local deviations in the point cloud. We implemented algorithms to flag areas where deviations exceed a threshold, indicating potential flaws in the casting part. This extends the system’s utility beyond mere metrology to quality assurance, providing a holistic view of casting part integrity.
Conclusion and Future Directions
In conclusion, our 3D laser scanning-based dimensional measurement system represents a significant advancement in the inspection of aerospace casting parts. By integrating advanced scanning technology with robotics and automation, we have addressed the shortcomings of traditional methods, achieving high coverage, efficiency, and accuracy. From my first-hand experience, I can attest to the system’s robustness and reliability in handling complex casting parts under production conditions. The modular design allows for easy maintenance and upgrades, ensuring long-term viability in manufacturing environments.
The key achievements include a coverage rate of 98.1% for a representative casting part, dimensional measurements within tight tolerances, and full automation that reduces labor intensity. The use of formulas and tables in this article underscores the technical rigor behind the system, from error modeling to performance evaluation. Repeated emphasis on casting part throughout the discussion highlights the focus on this critical component in aerospace engineering.
Looking ahead, we plan to enhance the system with artificial intelligence for real-time defect classification on casting parts. This would involve training neural networks on historical scan data to automatically identify and categorize anomalies. Additionally, we aim to improve scanning speed further by optimizing robot trajectories and implementing multi-sensor fusion. Another avenue is to integrate the system with digital twin platforms, enabling continuous monitoring of casting part quality throughout the product lifecycle.
In summary, this project demonstrates the transformative potential of 3D laser scanning for aerospace casting part inspection. As casting parts become more complex and performance demands increase, such automated systems will be indispensable for ensuring safety and reliability. I am confident that our work contributes to the broader adoption of smart manufacturing technologies in the aerospace sector, paving the way for more efficient and precise production of casting parts.