The post-processing of casting parts, which typically involves the removal of gates, risers, and parting line flashes, remains a significant bottleneck in modern foundries, especially for medium and large components. Traditional methods rely heavily on manual labor, leading to inconsistent quality, high labor intensity, and hazardous working environments filled with fumes and particulates. For critical railway components like side frames and bolsters, which can exceed 2.5 meters in length with gate thicknesses up to 150 mm, the demands are particularly severe. The inherent casting deviations, which can be as high as 10 mm, combined with the high viscosity and hardness of materials like Grade B+ steel, render conventional mechanical, laser, or wire-cut methods ineffective. This paper details the first-person development and implementation of a fully automated, adaptive flame cutting system designed to address these challenges for medium and large casting parts, enhancing precision, consistency, and operational safety.
The core challenge in automating the flame cutting of large casting parts lies in accommodating their substantial size, weight, and positional inaccuracies. A manual operator naturally adjusts the torch path based on visual feedback, a capability absent in pre-programmed robotic systems. Our solution centers on a synergistic integration of a heavy-duty positioning system, industrial robots, and real-time vision-based path correction. The system is architected around a dual-station, dual-robot configuration to maximize throughput. Two cutting stations allow for continuous operation: while one station is engaged in automated cutting, the other can be used for manual loading and unloading of casting parts. This setup effectively masks the non-value-added time associated with part handling. Two industrial six-axis robots, mounted on a common long-travel linear axis, service these two stations. Each robot is equipped with a custom flame cutting torch and a laser displacement sensor. The casting part itself is securely clamped onto a single-axis positioner capable of rotating the heavy component through ±200 degrees, presenting all necessary surfaces to the robot’s cutting tool. The key technical specifications of the developed system are consolidated in the table below.
| System Parameter | Specification / Performance Indicator |
|---|---|
| Primary Cutting Fuel | Enhanced Natural Gas |
| Ignition Method | Automatic External Pilot Ignition |
| Compatible Materials | Cast Iron, Cast Steel, Carbon Steel Casting Parts |
| Maximum Cutting Thickness | 300 mm |
| Positioner Payload Capacity | 1.5 Metric Tons |
| Positioner Rotation Range | ±200° |
| Robot Reach | 2538 mm |
| Linear Axis Effective Travel | 5042 mm |
| Cutting Residual Height (Edge Cuts) | ≤ 3 mm |
| Cutting Residual Height (Non-Edge Cuts) | ≤ 10 mm |
| Path Following Precision | ± 0.3 mm |
| Single-Piece Cycle Time (Example Bolster) | ~13 minutes |
System Architecture and Core Modules
The automated cell is enclosed within protective fencing with integrated safety light curtains. Its operation is governed by a centralized PLC-based control system. The major mechanical and functional modules include the Robot Module, the Positioner and Fixture System, and the Linear Travel Unit.
1. The Robot and Adaptive Cutting Module
Each robot module is a self-contained unit responsible for the precise delivery of the cutting flame. It consists of the robot manipulator (35 kg payload), a rigid mounting base, the flame cutting apparatus, a laser displacement sensor, and a collision detection sensor. The combined weight of the end-effector (torch and sensors) is 12 kg, well within the robot’s capacity, ensuring dynamic stability during high-speed motions.
The flame cutting apparatus is engineered for reliability and precise gas control. It features a specialized cutting torch, custom-designed nozzles optimized for minimal slag and residual height, an external pilot igniter, and a manifold of solenoid valves for gas control. The automatic ignition sequence is a critical safety and reliability feature. The process is governed by the following control logic, where the state of each solenoid valve (Oxygen for Preheating – \(V_{op}\), Cutting Oxygen – \(V_{oc}\), Enhanced Natural Gas – \(V_{gas}\), Pilot Igniter Gas – \(V_{pilot}\)) is a function of time \(t\) and system states:
- Pilot Ignition (t = t0): \(V_{pilot}(t) = ON\), \(V_{gas}(t) = OFF\), \(V_{op}(t) = OFF\), \(V_{oc}(t) = OFF\). The igniter is energized, creating a stable pilot flame.
- Torch Preheating (t = t1): \(V_{pilot}(t) = ON\), \(V_{gas}(t) = ON\), \(V_{op}(t) = ON\), \(V_{oc}(t) = OFF\). The main gas and preheat oxygen are ignited by the pilot flame, heating the casting parts’ surface to the kindling temperature.
- Cutting Phase (t = t2): \(V_{pilot}(t) = OFF\), \(V_{gas}(t) = ON\), \(V_{op}(t) = ON\), \(V_{oc}(t) = ON\). The high-pressure cutting oxygen stream is introduced, initiating the exothermic reaction and blowing away the molten material.
The cornerstone of adaptability for varying casting parts is the laser displacement sensor. Before each cut, the robot executes a search pattern, moving the sensor over the expected location of the gate root. The sensor captures a dense point cloud profile of the surface. By fitting this data to a known geometric model (e.g., a line or circle), the system calculates the actual position \((X_a, Y_a, Z_a)\) and orientation of the cutting feature. The deviation from the programmed nominal path \((X_n, Y_n, Z_n)\) is computed as:
$$
\Delta P = \begin{bmatrix} \Delta X \\ \Delta Y \\ \Delta Z \end{bmatrix} = \begin{bmatrix} X_a \\ Y_a \\ Z_a \end{bmatrix} – \begin{bmatrix} X_n \\ Y_n \\ Z_n \end{bmatrix}
$$
This offset \(\Delta P\) is then dynamically applied to the robot’s cutting path program in real-time, enabling precise cuts despite the dimensional variances inherent in sand-cast large casting parts. The collision sensor acts as a final safeguard, triggering an immediate emergency stop if an unplanned contact force is detected.
2. The Positioner and Fixturing System
Handling heavy, irregularly shaped casting parts requires a robust positioning solution. The single-axis positioner is designed with a driven headstock and a passive tailstock. The headstock incorporates a servo motor coupled with a high-ratio precision reducer to provide smooth and accurate rotational movement. The repeatability of the positioner is critical for multi-pass operations and is specified at ±0.1 mm at a radius of 500 mm. The fixturing system uses pneumatic clamps with mechanical hard stops. The clamps provide both vertical and lateral holding forces, and their design is modular to accommodate different families of casting parts, such as various bolster and side frame configurations, with quick changeover. The clamping force \(F_c\) must satisfy the condition to withstand the dynamic forces during rotation and cutting \(F_{dynamic}\):
$$
F_c \cdot \mu > F_{dynamic} = m \cdot a + F_{torch}
$$
where \(\mu\) is the coefficient of friction between the clamp and the casting parts, \(m\) is the mass of the component, \(a\) is the maximum acceleration, and \(F_{torch}\) is the negligible reaction force from the cutting process.
3. The Linear Travel Unit and System Integration
The linear travel unit forms the backbone of the dual-station concept. It is a heavy-duty structure comprising ground-mounted hardened steel rails, a precision gear rack, and a moving platform (skid) that carries the robot modules. A servo motor on the skid drives a pinion gear, propelling the entire robot assembly along the rack with a repeatability of ±0.1 mm. This allows a single robot to service multiple workstations. In our configuration, the two robots are fixed on a common skid, ensuring they maintain perfect relative positioning. The control architecture integrates the positioner and the linear axis as external axes of the robot controller. This allows for coordinated motion, where the robot’s cutting path and the part’s rotation are perfectly synchronized in a single programmed coordinate space, which is essential for complex contours on large casting parts.
Process Validation and Quantitative Analysis
The system was installed and subjected to extensive trials on production-grade railway bolsters. The process flow is fully automated: part loading, automatic travel to the cutting station, sequential scanning and cutting of all gates on one face, rotation to the next face, and final return to the load/unload station.
The performance was quantitatively evaluated against manual cutting, the incumbent method. The primary metrics were cutting surface quality and residual height (the remaining stub of the gate). Visual inspection confirmed that the automated system produced consistently smooth cut surfaces with minimal, easily removable slag. Manual cuts, in contrast, were often irregular and left significant, uneven stubs. A statistical comparison was performed by measuring the residual height on multiple gates from two automatically cut and two manually cut casting parts.
| Cutting Method | Average Residual Height – Edge Cuts (mm) | Standard Deviation – Edge Cuts (mm) | Average Residual Height – Non-Edge Cuts (mm) | Standard Deviation – Non-Edge Cuts (mm) |
|---|---|---|---|---|
| Automated Flame Cutting | 2.06 | 0.45 | 7.50 | 1.12 |
| Manual Flame Cutting | 10.06 | 3.87 | 14.00 | 4.25 |
The data unequivocally demonstrates the superiority of the automated system. The average residual height was reduced by approximately 80% for edge cuts and 46% for non-edge cuts. More importantly, the standard deviation was drastically lower for the automated process, indicating exceptional consistency and predictability—a key advantage for downstream grinding operations when processing these casting parts.
The cutting process itself can be modeled to understand the relationship between parameters. The cutting speed \(v\) must be optimized for material thickness \(d\). An empirical relationship derived from testing on B+ steel casting parts can be approximated by:
$$
v(d) = \frac{K}{d^{\alpha}}
$$
where \(K\) is a constant dependent on gas composition and pressure, and \(\alpha\) is an exponent typically between 0.5 and 1 for oxy-fuel cutting. For our system with enhanced natural gas, optimal results for casting parts with thicknesses between 30 mm and 120 mm were achieved with \(\alpha \approx 0.7\). The total cycle time \(T_{total}\) for a part is the sum of positioning, scanning, cutting, and idle times:
$$
T_{total} = \sum_{i=1}^{n} (t_{pos,i} + t_{scan,i} + \frac{L_i}{v(d_i)}) + t_{idle}
$$
where \(n\) is the number of gates, \(L_i\) and \(d_i\) are the length and thickness of the i-th gate, and \(t_{pos,i}\), \(t_{scan,i}\) are the associated positioning and scanning times. For the tested bolster with 14 gates of varying sizes, \(T_{total}\) averaged 13 minutes, representing a significant productivity gain per operator compared to manual methods.
Conclusions and Technological Impact
The development and successful implementation of this automated flame cutting system mark a substantial advancement in the post-processing of medium and large casting parts. By integrating industrial robotics, coordinated external axes, and real-time vision-based adaptive control, the system effectively solves the core problem of dealing with the significant dimensional variability of sand-cast components. It successfully replaces a labor-intensive, skilled, and hazardous manual operation with a consistent, precise, and programmable automated process.
The quantitative results prove its efficacy: it meets and exceeds the target specifications for cutting precision (≤ ±0.3 mm), residual height (≤3 mm for edges), and most importantly, consistency. The reduction in residual height directly translates to reduced time and cost in subsequent grinding operations for these massive casting parts. Furthermore, the dual-station, dual-robot design ensures high equipment utilization and production throughput. This system provides a scalable and flexible automation template not only for railway foundries but for any industry dealing with the challenging post-processing of large, high-value casting parts, paving the way for smarter, safer, and more efficient manufacturing environments.