⚡ electrical engineering

Flight Parameter Identification System Based on Cone-Shaped Piezoelectric Cantilever Array

This study proposes and validates a novel intelligent aircraft head model system integrating a cone-shaped piezoelectric cantilever array with a convolutional neural network, demonstrating superior accuracy in predicting airspeed, angle of attack, and angle of sideslip compared to state-of-the-art methods.

Original authors: Yangyang Huang, Fang Liu, Jie Hu

Published 2026-10-05

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Yangyang Huang, Fang Liu, Jie Hu

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

To steer a modern aircraft, especially a military one, pilots and computers need to know three things instantly: how fast the plane is moving through the air, the angle at which the nose is pointing relative to the wind, and the angle at which the wind is hitting the side of the fuselage. For decades, engineers have relied on physical tubes sticking out of the wings to measure these forces. While effective, these tubes add weight, create drag that slows the plane down, and can fail if the electronic signals they send are jammed by interference. In recent years, scientists have looked toward "smart skins"—surfaces that can feel the wind without needing bulky external parts. These skins often use materials that generate electricity when they bend or stretch, turning the physical push of the air into a signal the computer can read. The challenge has been making these smart skins accurate enough to replace the old, heavy tubes, particularly when the plane is banking or turning sharply.

A team of researchers has developed a new way to solve this problem by creating a smart aircraft nose that feels the wind using a ring of tiny, flexible sensors. Instead of sticking a tube out into the air, they built a cone-shaped model of an aircraft head and embedded a series of piezoelectric cantilevers into its surface. These cantilevers are essentially tiny, stiff strips of a special material that produces a voltage when it vibrates. The researchers arranged four of these strips in a circle around the cone, with each strip pointing in a different direction. When air flows over the cone, it pushes against these strips, causing them to vibrate. The faster the air moves or the more the wind hits the cone from the side, the more the strips bend and vibrate, generating a specific electrical signal. To make sense of these complex, wiggling signals, the team did not rely on simple math formulas alone. Instead, they trained a type of artificial intelligence, known as a convolutional neural network, to listen to the electrical output from the four strips and instantly calculate the speed of the wind and the angles of the aircraft's movement.

The researchers tested this system in a wind tunnel, a facility that simulates airflow by blowing air over a stationary object. They set the airspeed to various levels, ranging from a gentle breeze up to speeds of 30 meters per second, and they tilted the model to simulate different angles of attack and sideslip, turning it from minus 20 degrees to plus 20 degrees. As the air rushed over the cone, the sensors began to vibrate. The team found that the vibration of the strips was directly linked to the force of the wind. When the airspeed increased, the strips vibrated with greater amplitude, and the electrical voltage they produced rose accordingly. The shape of the cone itself played a crucial role; the researchers noted that the cone's design acted somewhat like a resonant chamber, amplifying the vibrations so that the sensors could detect even subtle changes in the airflow. This amplification allowed the system to be highly sensitive without needing massive, heavy components.

To ensure the system could handle the complex reality of flight, where wind speed and direction change simultaneously, the researchers used their artificial intelligence algorithm to interpret the data. They fed the electrical signals from the four sensors into the computer program, which had been trained on thousands of examples of how the sensors behaved under different conditions. The results were remarkably precise. When the system predicted the airspeed, the angle of attack, and the angle of sideslip, the errors were incredibly small. The difference between what the system predicted and what the actual wind tunnel instruments measured was less than 0.13 meters per second for speed, less than 0.15 degrees for the angle of attack, and less than 0.12 degrees for the sideslip angle. These margins of error are significantly smaller than those achieved by previous methods using different types of smart sensors, such as those based on heat or pressure changes.

The study also revealed the limits of the design. The researchers discovered that the sensors worked best when the wind hit the front of the cone within a specific range. If the wind came from too far to the side, beyond 20 degrees, the airflow became chaotic and created swirling vortices that confused the sensors, causing the electrical signals to become erratic. This finding helped the team define the optimal operating range for their device, confirming that it is most reliable when the aircraft is flying relatively straight or making moderate turns. By combining a clever physical design with a powerful learning algorithm, the team demonstrated that a smooth, cone-shaped surface covered in tiny vibrating strips can measure flight parameters with a level of accuracy that rivals or exceeds traditional instruments. This approach suggests a future where aircraft could be covered in a seamless, intelligent skin that feels the wind, offering a lighter, faster, and more reliable way to navigate the skies.

Technical Summary: Flight Parameter Identification System Based on Cone-Shaped Piezoelectric Cantilever Array

Problem Statement
Accurate measurement of flight parameters—specifically airspeed ( $v_{air}$ ), angle of attack ( $\psi$ ), and angle of sideslip ( $\gamma$ )—is critical for the attitude control of advanced flight vehicles. Traditional methods utilizing pitot tubes (PTs), multiple pressure sensing ports (MPSPs), and Global Navigation Satellite Systems (GNSS) face significant limitations. PTs and MPSPs increase aircraft weight, affect flight performance, and suffer from signal lag due to long pipeline distances, compromising real-time responsiveness. Furthermore, GNSS is vulnerable to electromagnetic interference from microwave weapons. While Smart Materials and Structures (SMS) such as capacitive, piezoresistive, and thermal flow sensors offer lightweight and conformable alternatives, existing systems often exhibit insufficient prediction accuracy for the rapid attitude response required by military aircraft. Additionally, there is a lack of reported systems that simultaneously fuse aircraft head surface vibration data with Convolutional Neural Network (CNN) algorithms to estimate all three parameters ( $v_{air}$ , $\psi$ , and $\gamma$ ) with high precision.

Methodology
The authors propose a novel intelligent Aircraft Head Model (AHM) system integrated with a Piezoelectric Cantilever Array (PCA) and a CNN-based prediction algorithm.

Hardware Design (AHM-PCA):
- Structure: A cone-shaped frame (CSF) made of ABS features a novel ring configuration of rectangular slots. Four piezoelectric bimorph cantilevers (PBCs), constructed from PZT-5 material, are installed in these grooves. The PBCs are arranged with a 90° projection angle in the y-z plane.
- Dimensions: The cone is 120 mm long with an 80 mm diameter and a 40° conical angle. Individual PBCs measure 36 mm in length, 2 mm in width, and 0.2 mm in thickness.
- Electronics: The PBCs utilize a parallel electrode connection model. The system includes a charge amplifier, data acquisition card, and a smart wind tunnel capable of generating airflow from 0 to 40 m/s.
Theoretical Modeling:
- A mathematical model for the PBC vibration mechanism was established using electromechanical coupling equations solved via MATLAB.
- The model relates the vibration displacement $y(t)$ to air pressure $P(t)$ and the response voltage $U(t)$ . It incorporates the mechanical damping ratio, resonant frequency, and electromechanical coupling terms.
- The analysis suggests the cone-shaped frame acts similarly to a Helmholtz resonator, enhancing flow pressure and amplifying the PBC vibration amplitude.
Algorithm (CNN):
- A Convolutional Neural Network was employed to map the multi-input sensor signals (response voltages from four PBCs) to the multi-output flight parameters ( $v_{air}$ , $\psi$ , $\gamma$ ).
- The network utilizes convolutional layers to extract spatial/temporal features, pooling layers to reduce dimensionality, and fully connected layers for final prediction.
- A composite loss function combining Mean Squared Error (MSE) for each parameter was used for training.
Experimental Validation:
- Scheme 1: Characterization experiments measured response voltage and vibration amplitude across airspeeds (0–30 m/s) and angles ( $\psi, \gamma$ from -40° to 40°).
- Scheme 2: A parameterization experiment was conducted to train and validate the CNN. Airspeeds ranged from 0 to 28 m/s, while $\psi$ and $\gamma$ were varied between -20° and 20°.

Key Results

Vibration Characteristics: Experimental results confirmed a quadratic positive correlation between airspeed ( $v_{air}$ ) and both the root mean square (RMS) response voltage and vibration amplitude. The analytical model showed high consistency with experimental data.
Sensing Characteristics: The system demonstrated that the response voltage of the PBCs approximates a quadratic function with respect to $\psi$ and $\gamma$ when the angles are less than 20°. However, within the specific range of -20° to 20°, the experimental results indicate linear sensing characteristics for both $\psi$ and $\gamma$ . Outside this range (e.g., beyond ±20°), flow field instability and vortex shedding cause chaotic signal increases, defining the effective operational limits.
Prediction Accuracy: The CNN algorithm successfully predicted flight parameters with the following absolute errors:
- Airspeed ( $v_{air}$ ): < 0.13 m/s
- Angle of Attack ( $\psi$ ): < 0.15°
- Angle of Sideslip ( $\gamma$ ): < 0.12°
Response Time: The system exhibited a response time of approximately 0.6 seconds.

Significance and Claims
The paper claims that the proposed AHM-PCA system, combined with the CNN algorithm, achieves superior accuracy compared to previous State-of-the-Art (SOTA) methods utilizing Smart Materials and Structures. The authors quantify these improvements as follows:

Airspeed Accuracy: 208% improvement (0.13 m/s error vs. 0.27 m/s in prior SOTA).
Angle of Attack Accuracy: 580% improvement (0.15° error vs. 0.87° in prior SOTA).
Angle of Sideslip Accuracy: 225% improvement (0.12° error vs. 0.27° in prior SOTA).

The authors conclude that this approach provides a new pathway for achieving fast-response, miniaturized, and electromagnetic-interference-free flight parameter identification. They suggest the device has the potential to be further miniaturized to the millimeter level for use as intelligent skin on advanced flight vehicles.

More like this