1. The Source-Path-Receiver Paradigm in NVH Engineering
The primary methodology for diagnosing and resolving complex noise, vibration, and harshness (NVH) issues in automotive, aerospace, and industrial machinery is the Source-Path-Receiver paradigm. This system-level framework breaks down any acoustic or tactile annoyance into three distinct parts:
1.
Source: The components that generate vibrational or acoustic energy through physical processes (e.g., engine combustion, electric motor electromagnetic torque ripple, air compressor piston reciprocating inertia, tire-road contact friction).
2.
Path: The physical structures or acoustic media through which the generated energy propagates from the source to the receiver.
3.
Receiver: The target location where human occupants experience the noise or vibration (e.g., driver\'s ear cabin sound pressure, driver\'s hands steering wheel vibration, occupants\' feet seat rail floor vibration).
This paradigm is illustrated below:
Airborne vs. Structure-Borne Classification
Path analysis differentiates between the physical media through which waves propagate:
โข
Structure-Borne Path: The source vibrates and transmits dynamic forces through its mechanical attachment points (mounts, bushings, links) into the receiving structural body. This structural vibration propagates through the body panels, which then act as large acoustic speakers (re-radiation) radiating noise into the cabin, or directly excite tactile response points.
โข
Airborne Path: The vibrating surface of the source directly radiates acoustic waves into the surrounding air cavity. This airborne sound energy travels to the exterior cabin walls, panel-vibrates the firewall, windshield, or doors, and is transmitted into the cabin cavity.
1.2 Transfer Path Analysis (TPA) Definition
Transfer Path Analysis (TPA) is a test-based or combined test/CAE engineering method used to quantify and rank the individual contributions of various structure-borne and airborne transmission paths to the receiver response.
By mapping the system, an engineer can answer questions such as:
1.
*Is the cabin boom noise caused by the left engine mount or the exhaust hanger?*
2.
*Is it airborne sound leaking through the door seals, or structure-borne force acting on the suspension links?*
3.
*How much load reduction at the engine mount is required to meet the cabin SPL target of 65 dBA?*
2. Force Identification Methods
The primary challenge of TPA is the identification of operational forces ${F}$ acting at the source attachment points. These forces cannot be measured directly using standard load cells during vehicle operation, as installing load cells alters the local structural stiffness and boundary conditions of the mounts. Therefore, we use indirect measurement techniques.
2.1 The Mount Stiffness Method (Direct Method)
The mount stiffness method computes the operational forces ${F_i}$ acting through elastomeric mounts by measuring the relative displacement across the mount under operating conditions.
For an elastomeric mount $i$ acting along a specific translational axis:
$X_{active,i}(omega)$ is the complex frequency-domain displacement measured on the source (active) side of the mount.
โข
$X_{passive,i}(omega)$ is the complex displacement measured on the body (passive) side of the mount.
โข
$K_i^*(omega) = K'_i(omega) + i K''_i(omega) = K'_i(omega) [1 + i \eta_i(omega)]$ is the frequency-dependent complex dynamic stiffness of the elastomer, where $K'$ is the storage stiffness and $\eta$ is the loss factor.
Engineering Requirements and Pitfalls:
1.
Dynamic Stiffness Calibration: The static stiffness of rubber is significantly lower than its high-frequency dynamic stiffness. The complex dynamic stiffness $K_i^*(omega)$ must be pre-measured in a specialized laboratory test rig under pre-load, temperature, and amplitude conditions matching the operating state.
2.
Phase Accuracy: Accelerometers must be placed extremely close to the mount brackets to avoid local panel resonances from introducing phase shifts. Displacement is obtained from acceleration by double integration in the frequency domain ($X(omega) = -a(omega)/\omega^2$).
3.
Cross-Axis Coupling: Real mounts transmit forces in three translation and three rotational axes. Ignoring rotational stiffness and cross-axis terms introduces significant errors at frequencies above $200\text{ Hz}$.
2.2 The Matrix Inversion Method (Indirect Method)
When mounts cannot be easily instrumented or their dynamic stiffness properties are unavailable, we use the Matrix Inversion Method. This method treats the passive structure as a multi-directional force transducer.
Step 1: Calibration (Offline Test): The source is disconnected or uncoupled. We measure a matrix of transfer functions (FRFs) $[H_{ax}]$ between the connection points (where operating forces act) and a set of indicator acceleration response points on the passive structure:
$$\{a_{ind}\} = [H_{ax}] \{F\}$$
2.
Step 2: Operational Measurement: The system is run in its operating state. We measure the operating accelerations $\{a_{ind, ops}\}$ at the indicator points.
3.
Step 3: Force Reconstruction: The operational forces are identified by inverting the calibration FRF matrix:
$$\{F_{ops}\} = [H_{ax}]^{-1} \{a_{ind, ops}\}$$
Overdetermined System and Pseudo-Inverse:
To increase accuracy and averaging, we select the number of indicator response points $M$ to be larger than the number of unknown forces $N$ ($M > N$). This yields an overdetermined system. We solve for the forces by minimizing the sum of squared errors between the measured accelerations and the reconstructed predictions, which leads to the Moore-Penrose pseudo-inverse:
where $[H_{ax}]^H$ is the conjugate transpose of $[H_{ax}]$.
Matrix Ill-Conditioning and Regularization:
The inversion of $[H_{ax}]$ is highly susceptible to ill-conditioning, especially at frequencies near structural resonances or anti-resonances of the passive body. If the matrix has a high condition number:
โข
Small measurement noises in accelerations $\{a_{ind, ops}\}$ are magnified exponentially.
โข
The identified forces show huge, non-physical spikes.
To stabilize the force reconstruction, we apply Singular Value Decomposition (SVD) and Tikhonov Regularization (see Derivation section).
3. Path Ranking and Sound Contribution
Once the operational forces $\{F_{ops}\}$ have been determined (using either the Mount Stiffness or Matrix Inversion method), the final step of TPA is to calculate the individual path contributions to the target receiver.
For a target acoustic receiver point (e.g., driver\'s ear cabin sound pressure $P_{target}(omega)$), we measure the vibro-acoustic transfer functions (FRFs) $[H_{pf}]$ from the connection points to the receiver. The contribution of structure-borne path $j$ is:
$$P_j(omega) = H_{pj}(omega) \cdot F_j(omega)$$
The total acoustic response is the complex vector sum of all path contributions:
A common mistake in path contribution analysis is summing the absolute magnitudes of the paths. Because $P_j(omega)$ is a complex number ($P_j = |P_j| e^{i\phi_j}$), the phase angle $\phi_j$ plays a critical role:
โข
In-Phase Paths (Constructive Interference): If two paths have similar phases (e.g., $\phi_1 \approx \phi_2$), their sound pressures add constructively.
โข
Out-of-Phase Paths (Destructive Interference): If two paths are out of phase ($\phi_1 \approx \phi_2 + 180^\circ$), they cancel each other out.
Case Study Danger:
Suppose Path 1 contributes $+0.5\text{ Pa}$ and Path 2 contributes $-0.4\text{ Pa}$. The net pressure is $0.1\text{ Pa}$. If an engineer, seeing that Path 1 is the largest contributor, stiffens the bracket for Path 1 and reduces its contribution to $0.0\text{ Pa}$, the new net pressure becomes $-0.4\text{ Pa}$ (which corresponds to an increase in sound pressure level from $74\text{ dB}$ to $86\text{ dB}$).
Thus, path ranking must always be performed using vector contributions rather than scalar magnitudes.