⚡ electrical engineering

Sensor-Limited Observability and Carrier-Induced Reachability of Low-Order Rotor-Coupled NVH in Production Electric Drives: A Magnetic Co-Energy, Gramian, and Active Projection Framework for Production-Signal Feasibility Analysis

This paper proposes a magnetic co-energy and Gramian-based framework demonstrating that while low-order rotor-coupled NVH in production electric drives is often unobservable and unreachable via standard passive signals due to flux-linkage sensitivity, it can be made detectable and controllable through a novel carrier-induced active projection strategy.

Original authors: Meng-Chou Wu

Published 2026-07-28

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Meng-Chou Wu

Original paper licensed under CC BY 4.0 (http://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

The Invisible Rumble: When Motors Whisper What Sensors Can't Hear

Imagine you are listening to a symphony orchestra. Most of the time, you hear the violins and the trumpets clearly. But sometimes, a single cello in the back row starts to vibrate a low, uncomfortable hum. If that cello is hidden behind a thick velvet curtain, your ears might not pick up the sound, even though the instrument is shaking violently. In the world of electric cars and high-speed motors, this "curtain" is a major problem. These motors are incredibly complex machines that spin at thousands of revolutions per minute. Inside them, the metal parts can vibrate in specific patterns, creating noise and wear that engineers call NVH (Noise, Vibration, and Harshness).

To fix these problems, engineers usually rely on sensors. They stick accelerometers on the motor or listen to the electrical signals flowing through the wires, hoping to hear the "cough" of a vibrating part. But here is the catch: modern electric motors are so tightly packed and so fast that the standard sensors often miss the most dangerous vibrations. It's like trying to hear a whisper in a hurricane. The big question for scientists is: If the motor is shaking in a way that our standard tools can't see, can we trick the motor into revealing its secrets? Can we send in a special "probe" signal that makes the invisible vibration become visible, or even stop it? This is the puzzle this paper tackles, using a mix of magnetic math and clever signal tricks.

The Paper's Story: Hunting the Ghost Vibration

This paper is about a specific, annoying problem found in real-world electric drives: a low-order vibration (specifically a "Order-2" mode, which means the motor deforms in a shape with two bumps) that happens when the motor is under load. This vibration is like a ghost; it's physically strong enough to cause noise and wear out the motor, but it is "invisible" to the standard electrical sensors that the car's computer uses. The computer sees the current and voltage, but it can't "see" the air gap between the spinning rotor and the stationary stator changing shape.

The authors, led by Meng-Chou Wu, set out to answer a simple but tricky question: Can we detect or control this ghost vibration using only the signals the car already has? They didn't just guess; they built a mathematical model to see if the physics even allows it.

The "Invisible" Problem

The researchers found that for many common motor designs, this specific vibration is mathematically "orthogonal" to the electrical signals. To use an analogy, imagine trying to measure the wind by looking at a shadow cast by a tree. If the wind is blowing exactly parallel to the ground, the shadow doesn't change length, even though the wind is strong. Similarly, the motor's electrical currents (the $I_d$ and $I_q$ signals) are like that shadow. The vibration (the wind) is happening, but because of the motor's perfect symmetry, the electrical signals don't change at all. The authors call this "passive degeneracy."

They proved that under normal, passive conditions (just listening to the motor without doing anything special), the standard electrical signals and the rotor position sensor (the resolver) are often blind to this specific type of vibration. The math showed that the "flux-linkage sensitivity vector" (a fancy way of saying "how much the magnetic field changes when the gap changes") is effectively zero for these low-order modes. So, if you just sit there and listen, the ghost remains a ghost.

The "Magic Carrier" Solution

But the paper doesn't just say "it's impossible" and walk away. It proposes a clever workaround: Active Carrier Injection.

Think of the motor like a dark room where you can't see the furniture. If you just stand there, you see nothing. But if you shine a flashlight (the "carrier") that sweeps back and forth, the shadows of the furniture start to dance. The paper suggests injecting a specific electromagnetic signal—a "carrier"—into the motor. This isn't just any signal; it's a high-frequency wave designed to interact with the motor's magnetic field.

When this carrier signal mixes with the motor's main magnetic field, it creates a "beat" pattern. If the carrier is tuned correctly (specifically, if its order is the motor's pole pairs plus or minus the vibration order), it acts like a spotlight. Suddenly, the invisible vibration modulates the carrier signal. Instead of the vibration being invisible, it creates "sidebands"—new, detectable ripples in the electrical signal that appear at specific frequencies ( $h \pm r$ ).

The authors call this "carrier-induced observability." They show that even if the vibration is invisible to the normal sensors, it becomes visible when the motor is "probed" with this special carrier. It's like the vibration was hiding in the dark, but the carrier signal forced it to wear a neon vest that the sensors can finally see.

The Catch: It's Not a Magic Wand

The paper is very careful not to overpromise. They ran two types of simulations to test their ideas, and the results are nuanced:

The Passive Test: They simulated a motor with 48 slots and 8 poles. Even with a realistic model of the motor's winding, the "ghost" vibration remained invisible to passive signals. This confirms that you can't just rely on the existing data to fix the problem.
The Carrier Test: They simulated injecting a carrier. They found that for the vibration to be detected, the "sideband" signal created by the carrier must be strong enough to rise above the background noise. They calculated that the signal strength needs to be about 3 to 5 times higher than the noise floor to be reliably detected. If the noise is too loud or the carrier is too weak, the ghost stays hidden.

The paper also points out a crucial distinction: Seeing is not the same as Touching. Just because the carrier makes the vibration visible (observable) doesn't mean the carrier can stop it (reachable). To stop the vibration, the carrier must also be able to push back against the magnetic forces causing the shake. The authors show that while the math allows for this "reachability," it depends on specific conditions that might not always be met in a real, noisy, imperfect motor.

What This Means for the Future

The authors are clear that this is a framework, not a finished product. They haven't built a car that fixes this problem yet. Instead, they have built a map. They have shown where the problem lies (the invisible gap) and how to look for it (the carrier trick).

They outline a path forward:

First, we need better magnetic maps (like a detailed blueprint of the motor's insides) to know exactly how the carrier will behave.
Second, we need to prove that the motor's electronics can actually handle the extra "probe" signal without getting too hot or running out of power.
Finally, we need to test this on real hardware with microphones and vibration sensors to see if the "neon vest" really works in the real world.

In short, this paper tells us that while some motor vibrations are invisible to our standard tools, they aren't necessarily untouchable. By using a clever "flashlight" signal, we might be able to find and fix these ghosts, but it requires careful engineering and more testing to make sure the flashlight doesn't just blind us or burn out the motor. The door is open, but we still have to walk through it.

Technical Summary: Sensor-Limited Observability and Carrier-Induced Reachability of Low-Order Rotor-Coupled NVH in Production Electric Drives

Problem Statement
This paper addresses a practical Noise, Vibration, and Harshness (NVH) problem observed in production electric drives: low-order rotor-coupled vibration components (specifically Order-2, but generalized to orders $r \in \{1, 2, 3, 4\}$ ) that contribute to acoustic discomfort and durability risks. In high-speed traction motors, mechanical mitigation is often impractical due to compact rotor structures, strict balance requirements, and fixed electromagnetic constraints.

The core feasibility challenge arises from sensor limitations in production controllers. These systems typically lack direct measurements of rotor vibration or housing acceleration. Instead, they rely on electrical signals ( $I_d, I_q$ , voltage commands) and resolver signals (angle $\theta_r$ , speed $\omega_r$ ). The paper investigates whether a mechanically significant low-order air-gap deformation can be inferred (observed) or influenced (reached) using only these production-accessible signals.

Methodology
The authors formulate the problem using a control-oriented electromechanical model based on magnetic co-energy.

Modeling: The selected rotor-coupled vibration is represented as an equivalent air-gap deformation mode with state $x_g = [g, v_g]^T$ , where $g$ is the generalized deformation coordinate and $v_g$ is its modal velocity. The system is modeled as a six-state nonlinear system ( $I_d, I_q, \theta_r, \omega_r, g, v_g$ ) linearized around a fixed-speed, fixed-current loaded operating trajectory.
Passive Analysis: The paper employs finite-time Gramians to analyze the passive linearized plant. Three specific measures are defined:
- Current-based Observability: Can $g$ and $v_g$ be inferred from $I_d/I_q$ residuals?
- Resolver-based Observability: Can they be inferred from $\theta_r$ and $\omega_r$ ?
- Voltage Reachability: Can $V_d/V_q$ commands influence $g$ and $v_g$ ?
Active Analysis: Recognizing that passive signals may be insufficient, the paper introduces a "carrier-induced projection" framework. This involves injecting a physically realizable electromagnetic carrier to create carrier-dependent measurement ( $\Gamma_y(m)$ ) and force ( $\Gamma_F(m)$ ) projections, potentially rendering invisible modes observable or reachable.
Numerical Studies: Two physically anchored numerical feasibility studies are conducted to demonstrate the framework's computational structure:
- An architecture-aware passive projection study using a 48-slot, 8-pole symmetry proxy.
- A carrier-induced residual detectability study evaluating sideband Signal-to-Noise Ratio (SNR) under noise and near-order disturbance leakage.

Key Contributions and Findings

Passive Degeneracy: The analysis identifies the flux-linkage sensitivity vector $\ell_\lambda = [\bar{\lambda}_{d,g}, \bar{\lambda}_{q,g}]^T$ as the common physical bottleneck for both passive current observability and passive voltage reachability. Under ideal symmetric conditions, low-order air-gap deformations ( $0 < r < p$ ) are often orthogonal to the dq electromagnetic projection, resulting in $\ell_\lambda \approx 0$ . Consequently, these modes can be mechanically significant yet "first-order invisible" to production electrical signals.
Resolver Limitations: Resolver observability is governed by a separate torque-projection margin ( $\rho_T$ ). Even if $\ell_\lambda \neq 0$ , the deformation may not produce a usable resolver signature if the resulting torque perturbation is attenuated by rotor inertia or estimator filtering.
Carrier-Induced Projection: The paper demonstrates that passive invisibility ( $\ell_\lambda \approx 0$ ) does not imply carrier-induced invisibility. By injecting a carrier of spatial order $m$ (where $m = p \pm r$ ), the system can generate carrier-dependent coefficients $\Gamma_y(m)$ and $\Gamma_F(m)$ . These coefficients can create new measurement paths (via sidebands in residuals) and force paths (via generalized electromagnetic forces) that bypass the passive degeneracy.
Distinction of Feasibility: The framework explicitly separates observability (detection) from reachability (actuation). A carrier may make a mode observable without making it reachable, and vice versa. Closed-loop attenuation requires both conditions to be met.
Numerical Validation: The numerical studies confirm that under symmetric conditions, low-order passive projections collapse to the numerical floor. However, introducing low-order symmetry breaking (e.g., manufacturing variations) can reopen these paths. Furthermore, the detectability study shows that carrier-induced sidebands require a composite projection strength ( $\epsilon A_g \|\Gamma_y\|$ ) to exceed the residual noise floor and must be distinguished from near-order disturbance leakage, which is mitigated by longer observation windows.

Significance and Claims
The paper claims to provide a feasibility framework rather than a complete solution or a validated production controller. Its primary significance lies in reframing the NVH problem from a control design task to a signal-structure analysis task. It establishes that:

Not all mechanically relevant NVH modes are accessible via standard production electrical signals due to spatial orthogonality.
Carrier injection is not merely a tool for sensorless position estimation or fault diagnosis but can serve as an active mechanism to alter the observability and reachability of specific deformation modes.
Before attempting active NVH suppression, engineers must verify the existence of specific projection paths ( $\ell_\lambda$ , $\rho_T$ , $\Gamma_y$ , $\Gamma_F$ ) and ensure that carrier-induced sidebands are distinguishable from noise and disturbances.

The authors explicitly state that the work does not constitute FEM-calibrated motor validation. The numerical studies are intended to demonstrate the computational structure and scaling behavior of the proposed framework. Future work, as outlined in the paper, requires FEM-calibrated magnetic maps, structural modal identification, carrier-on LTV/LTP simulations, and experimental validation with dynamometer testing to confirm the theoretical feasibility in real production hardware.