⚡ electrical engineering

Online Flatness Detection Method and Experimental Research of Aircraft Rudder Surface Based on Bidirectionally Coupled PSO-SA Hybrid Optimization Algorithm

This paper proposes a bidirectionally coupled PSO-SA hybrid optimization algorithm with adaptive inertia weights and noise-modified acceptance criteria to enable high-precision, efficient online flatness detection of aircraft rudder surfaces, achieving sub-micron system accuracy and significantly outperforming traditional methods in both convergence speed and fitting stability.

Original authors: Zeqing Yang, Jiayu Guan, Weiwei He, Yiding Yao, Yingshu Chen, Yanrui Zhang, Xuefei Zhang

Published 2026-07-15

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Zeqing Yang, Jiayu Guan, Weiwei He, Yiding Yao, Yingshu Chen, Yanrui Zhang, Xuefei Zhang

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to smooth out a crumpled piece of aluminum foil to make it perfectly flat, but this foil is actually a giant, curved airplane rudder. If it's not flat enough, the plane might wobble in the sky or even crash. The goal is to measure this "flatness" with incredible precision—down to the width of a human hair (micrometers)—while the part is still sitting on the factory machine, not in a separate lab.

The Problem: The "Local Trap"
Traditionally, factories used old-school methods to check this flatness, like using a dial gauge or a slow, offline machine. These methods were like trying to find the lowest point in a mountain range by only looking at the hill right in front of you. They often got stuck in "local traps," thinking a small dip was the deepest point, when actually, there was a much deeper valley nearby. They were also slow, easily confused by the tiny bumps and noise of the factory floor, and couldn't handle the complex curves of modern airplane parts.

The New Solution: A Team of Explorers
The researchers at Hebei University of Technology invented a new digital "brain" to solve this. They combined two famous search strategies into a single, super-smart team:

The Swarm (PSO): Imagine a flock of birds searching for food. They fly around together, sharing information about where the best spots are. This is great for scanning a huge area quickly.
The Heat Seeker (SA): Imagine a metal cooling down. As it cools, it settles into the most stable shape. This method is great for making tiny, careful adjustments to find the absolute best spot, even if it means taking a step backward first to avoid getting stuck.

The Big Twist: They Talk to Each Other
Here is the magic part. In older versions of these tools, the "Birds" would find a spot, hand it off to the "Heat Seeker," and then the Birds would go home and stop working. The Heat Seeker would do its thing, but the Birds never got to see the results.

The authors argue that this one-way street is a mistake. Their new method creates a two-way conversation. The Heat Seeker refines the answer and then sends it back to the Birds to update their map. This "bidirectional" loop means the whole team learns from every step, preventing them from getting stuck in those local traps.

Special Superpowers
To make this work on a noisy factory floor, they gave the algorithm two special upgrades:

The Adaptive Weight: The "Birds" don't just fly at a constant speed. If the terrain looks bumpy (noisy data), they adjust their flight pattern dynamically to explore more or focus tighter, depending on what the data is doing.
The Noise Filter: The factory machines make tiny, random vibrations. The new algorithm has a special rule that says, "If this weird jump looks like just a vibration, don't panic; keep searching." This stops the system from getting confused by false alarms.

The Proof: Simulations and Real Tests
The team didn't just guess this would work; they tested it in two ways:

In the Computer: They ran simulations on a computer (using MATLAB) comparing their new team against five other popular search methods. The results showed their new method was faster, more accurate, and much more stable. It found the "flattest" plane with an error of about 29.7 μm (micrometers), which was significantly better than the standard methods (which hovered around 42.5 μm and 48.3 μm).
In the Real World: They took 24 actual airplane rudder parts (some just the skin, some fully assembled) and measured them on a real CNC machine tool. After carefully calibrating their equipment to ensure the probe was perfect, they ran the tests.

The Results
The new method didn't just find a better answer; it did it faster.

Accuracy: The average flatness error was 29.7 μm. This is well within the strict aviation safety rule of 0.1–0.3 mm (which is 100–300 μm). The authors used statistical tests (ANOVA) to prove that their results were not just lucky guesses but were statistically significantly better than the other methods.
Speed: Checking one single rudder took only 2.1 minutes. This is a 34.4% reduction in time compared to the standard "Bird" method and a massive 63.2% reduction compared to the standard "Heat" method.

What They Say It's NOT
The authors are very clear about what their method is not. They explicitly state that simply pasting old algorithms together (like a standard "Bird" followed by a standard "Heat" without them talking) doesn't work well enough for these curved, noisy surfaces. They also rule out traditional linear math methods because they can't handle the complex, non-flat shapes of the rudder.

The Bottom Line
This paper suggests that by letting two different search strategies talk to each other and by teaching them to ignore factory noise, we can measure airplane parts much faster and more accurately than before. It's a proven, working system that fits right into the factory line, turning a slow, tricky process into a quick, reliable check that keeps planes safe.

Technical Summary: Online Flatness Detection of Aircraft Rudder Surfaces via Bidirectionally Coupled PSO-SA

Problem Statement
The manufacturing of aircraft control surfaces, specifically rudder surfaces, requires micron-level flatness precision to ensure aerodynamic performance and flight safety. Traditional detection methods face significant bottlenecks:

Hardware Limitations: Contact methods (dial indicators) suffer from low sampling density and manual dependency, while Coordinate Measuring Machines (CMMs) require offline, time-consuming processes unsuitable for batch production.
Algorithmic Limitations: Existing plane fitting algorithms (e.g., Least Squares, standard Minimum Zone) struggle with the non-convex, multi-extremum nature of curved rudder surfaces. They are prone to local optima, lack robustness against measurement noise inherent in CNC trigger-probe sampling, and fail to meet the dual demands of high efficiency and micron-level accuracy in online inspection scenarios.

Methodology
The study proposes an integrated online detection system combining CNC trigger-probe hardware with a novel Bidirectionally Coupled Particle Swarm Optimization–Simulated Annealing (PSO-SA) Hybrid Optimization Algorithm.

Hardware System:
- A trigger-based on-machine inspection system is established using high-precision probes (Renishaw/BLUM series) mounted on CNC spindles.
- The system employs a closed-loop workflow: path planning (avoiding collisions via nearest-neighbor algorithms), trigger signal acquisition, real-time coordinate recording, and data processing.
- A rigorous multi-dimensional calibration protocol is implemented, including probe geometric parameter calibration, coordinate system association, and trigger signal hysteresis error compensation, reducing systematic errors to within 1 μm.
Algorithmic Framework:
- Core Innovation: Unlike traditional serial PSO-SA frameworks that use unidirectional information flow (PSO output $\to$ SA refinement), this study introduces a bidirectionally closed-loop architecture. The refined solutions from SA are fed back to update the PSO global optimum, creating a mutual calibration mechanism.
- Innovation 1: Residual-Adaptive Nonlinear Inertia Weight: A dynamic inertia weight strategy is developed for the PSO module. It adjusts based on the fluctuation characteristics of fitting residuals, expanding the search space in early iterations and narrowing it for high-precision local exploitation in later stages.
- Innovation 2: Noise-Modified Metropolis Criterion: The SA module utilizes a modified acceptance criterion that incorporates a noise threshold correction term. This allows the algorithm to probabilistically accept suboptimal solutions that may arise from random trigger-probe noise, preventing premature convergence to local optima caused by noisy data points.
- Objective: The algorithm minimizes the gap between the maximum and minimum signed distances of sampling points to a candidate plane, adhering to the Minimum Zone criterion for flatness evaluation.

Key Contributions

Theoretical Advancement: The paper establishes a dedicated bidirectionally coupled PSO-SA theory specifically tailored for the non-convex minimum zone flatness evaluation of curved free-form surfaces, addressing the limitations of generic serial hybrids and standalone metaheuristics.
System Integration: It presents a complete engineering workflow from measurement path planning and probe calibration to intelligent plane fitting and error evaluation, realized on a CNC machine tool.
Algorithmic Robustness: The proposed method explicitly targets the anti-noise robustness required for on-machine trigger sampling, a factor often overlooked in standard optimization literature.

Experimental Results
The study validates the method through MATLAB simulations and physical experiments on 24 aircraft rudder workpieces (aluminum alloy skins and assembled riveted components).

Simulation Performance: Compared against Standard PSO, Standard SA, CLPSO, Adaptive Cooling SA, and traditional Fixed-Coupling PSO-SA, the proposed algorithm demonstrated superior convergence speed, stability, and accuracy. It achieved the lowest objective function values and avoided the stagnation observed in benchmark algorithms.
Physical Measurement Accuracy:
- The average flatness error calculated by the proposed algorithm was 29.7 μm.
- This represents a 30.1% reduction compared to Standard PSO (42.5 μm) and a 38.5% reduction compared to Standard SA (48.3 μm).
- The results fall well within the aviation tolerance specification of 0.1–0.3 mm.
- Statistical analysis (ANOVA with Tukey's post-hoc test) confirmed that the accuracy improvements are statistically significant ( $p < 0.05$ ).
Efficiency: The full detection cycle for a single workpiece was reduced to 2.1 minutes, a 34.4% improvement over Standard PSO and a 63.2% improvement over Standard SA, meeting the takt time requirements for mass production.

Significance and Claims
The authors claim that this research provides a targeted theoretical basis and a complete engineering implementation scheme for the intelligent flatness detection of aerospace curved thin-walled parts. The study asserts that by integrating bidirectional optimization feedback with noise-adaptive mechanisms, the proposed method overcomes the inherent defects of existing hybrid algorithms (premature convergence, poor noise robustness, and unidirectional iteration).

The work is positioned as a valuable technical reference for form and position error evaluation of irregular industrial components under noisy measurement conditions. It demonstrates that high-precision, micron-level online inspection is achievable in batch manufacturing scenarios, directly addressing the need for automated quality control in standardized intelligent aerospace manufacturing systems. The paper concludes that the method successfully balances global exploration and local exploitation, offering a robust solution for the specific challenges posed by curved rudder surfaces.

Technical Summary: Online Flatness Detection of Aircraft Rudder Surfaces via Bidirectionally Coupled PSO-SA

More like this