Holographic Sound Field Inspection Instrument
Compressor Condition Monitoring Solution

A Non-contact Intelligent Diagnosis Solution for Reciprocating Compressors
Based on Spherical Panoramic Acoustic Technology

I. Industry Background and Monitoring Pain Points

Reciprocating compressors are the core power equipment of industrial stations such as coalbed methane (CBM) gathering stations, serving as the “heart” that handles gas gathering, multi-stage boosting, and transmission. They rely on a crank-and-connecting-rod mechanism to drive the piston in periodic reciprocating motion, compressing gas through discontinuous volumetric variation. This physical mechanism inevitably brings strong vibration, high-decibel radiated noise, and high-frequency faults; moreover, under long-term high-pressure, high-load, and harsh conditions of dust and temperature difference, it readily gives rise to latent faults such as excessive discharge back pressure, lubricating oil film rupture, pneumatic component wear, and rotor imbalance—which not only increase reactive kinetic energy losses but may also culminate in safety accidents.

Traditional monitoring methods rely heavily on contact-type sensor networks, which suffer from three major limitations:

• Coverage blind zones—contact sensors must be deployed invasively at critical load-bearing points or on pipe walls; for large parallel multi-cylinder units, dead-zone-free coverage of the machine’s entire physical field is difficult to achieve.

• Difficult explosion-proof wiring—sensor cabling and power supply face enormous challenges inside gathering stations with extremely stringent explosion-proof requirements.

• High operation and maintenance cost—high-frequency mandatory maintenance and replacement of wear parts create continuous cost pressure, and retrofits require shutdown for construction.

II. Solution Overview: Non-contact Panoramic Acoustic Monitoring

Centered on spherical panoramic acoustic technology, the holographic sound field inspection instrument achieves a fully non-contact inverse reconstruction of the internal operating state of a reciprocating compressor—without altering existing process piping or equipment enclosures—by acquiring multi-dimensional acoustic physical quantities in three-dimensional space (spatial distribution of sound pressure level, sound intensity vector, and three-dimensional source localization). Compared with conventional contact-type monitoring, the core differences are as follows:

Technical route: spherical-array panoramic acquisition → spherical harmonic sound field reconstruction → acoustic–optical fusion visualization → global physical quantity acquisition → state inversion and intelligent diagnosis.

III. Core Technologies

3.1 Spherical Harmonic Transform: The Mathematical Foundation of Spherical Panoramic Acoustic Imaging

By virtue of its high degree of three-dimensional spatial symmetry, the spherical microphone array synchronously maps all radiated noise arriving from every direction within a single measurement cycle. Within the mathematical framework, the spatial sound pressure field can be expanded by the Spherical Harmonic Transform (SHT): in the received sound pressure model p(k)=A(k)S(k)+z(k), the directional steering matrix A(k) can be decomposed as A(k)=Y(Ω)B(kr)YH(Γ), where the frequency term is concentrated in the diagonal matrix B(kr) and the source-direction matrix YH(Γ) no longer contains a frequency component. This achieves complete frequency–direction decoupling, so that the cross-spectral matrix can be computed directly in the frequency domain with spatial spectrum smoothing—laying the foundation for localization and reconstruction of broadband sound sources.

3.2 Low-Frequency Ill-Posed Effect and Transformer Sound Field Reconstruction

When processing the characteristic noise of a compressor (such as low-frequency fundamental-frequency pulsation below 500 Hz), conventional spherical harmonic beamforming encounters severe physical limitations: as kr→0 in the low-frequency band, the modal strength bn(kr) plunges rapidly toward zero, so that solving the inverse matrix becomes severely ill-posed and the condition number tends to infinity—tiny ambient noise or array mismatch is then infinitely amplified. In addition, the limited aperture of a portable device’s array (radius approximately 0.15–0.3 m) results in an abnormally broad low-frequency main lobe and high side lobes, making it difficult to clearly separate closely packed gas storage tanks, cylinders, and valves.

To address this, the solution introduces a hierarchical multi-task Transformer (HMT-SH-Transformer) sound field reconstruction model. Taking the normalized spherical harmonic feature matrices of each frequency band as input, it employs a multi-head self-attention mechanism to adaptively weight different spherical harmonic orders, precisely excluding the interference of ill-conditioned orders caused by low-frequency spherical Bessel functions; it then layers grid labels by frequency band (low frequency → low resolution, high frequency → high resolution) and fuses the outputs of the branches, simultaneously improving the convergence of low-frequency ill-posed signals and the imaging resolution of high-frequency transient impulses.

3.3 Acoustic Holographic Reconstruction and State Inversion

Based on the Green’s function matrix, the inverse boundary value problem of the acoustic wave equation (Helmholtz equation) is solved, precisely back-propagating the captured sound field data to the equipment surface and computing the absolute sound pressure distribution, the normal vibration velocity, and the sound intensity spatial vector. Here, normal vibration velocity is the core bridge connecting the external sound field to the internal mechanical dynamics: combined with the compressor’s multi-rigid-body dynamics and thermodynamic coupling model, the external normal vibration velocity can be deeply mapped to the internal unsteady gas forces (indicator-diagram P–V cycle parameters), inversely deriving internal thermodynamic state deterioration or disruption of hydrodynamic pressure balance (e.g., concealed gas blow-by, piston ring fracture).

3.4 Acoustic–Optical Pixel-Level Fusion and Intelligent Diagnosis

The system fuses panoramic visual images with spatial acoustic energy distribution images at the pixel level to generate an intuitive panoramic acoustic mapping image. It further converts the three-dimensional radiated sound field into two-dimensional pseudo-color slices, extracts gray-level co-occurrence matrix (GLCM) texture parameters, and combines them with refined composite multiscale dispersion entropy (RCMDE) to form a hybrid feature vector. This vector is input into a deep belief network (DBN) / bidirectional gated recurrent unit (BiGRU), which automatically mines high-dimensional nonlinear feature correlations and precisely clusters healthy boundaries and mixed fault states without any manually specified physical threshold.

IV. Sound Generation Mechanisms and Monitoring of Core Components

The acoustic signal of a reciprocating compressor is a complex mixture in which multiple physical fields interweave—aerodynamics, fluid–structure interaction, rotor dynamics, and others. Panoramic acoustic imaging technology provides in-depth analysis of the sound generation mechanisms of the three core components:

V. Typical Fault Panoramic Acoustic Diagnosis Cases

5.1 Excessive Discharge Back Pressure and Downstream Flow-Path Blockage

5.2 Holographic Localization of Knocking from Poor Crosshead Lubrication

5.3 Targeted Holographic Diagnosis of Cylinder–Piston Seizure

VI. Field-Measured Data and Technical Specifications

Field measurement was carried out using a non-stop, no-retrofit inspection approach, completed in complex reverberant background noise above 90 dB at a gas gathering station. The measured sound pressure levels at typical locations are as follows:

Core product specifications:

VII. Application Value: From “Abnormal Sound Perception” to “Spatial Source Inversion”

VIII. Summary and Industrial Deployment

Taking “spherical harmonic transform + Transformer sound field reconstruction + acoustic holographic reconstruction + state inversion” as its core technical route, the holographic sound field inspection instrument provides reciprocating compressors with a non-contact, panoramic-coverage, quantitatively inverting intelligent condition monitoring solution, fundamentally overcoming the coverage blind zones, explosion-proof wiring difficulties, and operation and maintenance costs of conventional contact-type monitoring.

Industrial deployment roadmap:

• Lightweight on-device algorithm deployment—the diagnostic model is made lightweight to enable real-time on-site edge inference.

• Edge computing and global monitoring network—multiple units are networked to build a station-level global acoustic monitoring network.

• Digital twin integration and iteration of the O&M paradigm—acoustic data is connected to a global digital twin hub, driving the leap from “reactive post-failure repair” to “intelligent predictive condition-based maintenance.”