Anti-interference sound vibration detection method and device

Abstract

The invention provides an anti-interference sound and vibration detection method and device, and relates to the field of sound and vibration detection. The method comprises the following steps: acquiring target special equipment and an equipment type corresponding to the target special equipment; according to the equipment type, acquiring a multi-channel high-sensitivity sensing array combination corresponding to the equipment type from a preset inspection method database; acquiring sound vibration signal data corresponding to the target special equipment based on the multi-channel high-sensitivity sensing array combination; performing modal decomposition on the sound vibration signal data by applying an improved empirical mode decomposition algorithm, eliminating pseudo-modal components and reconstructing effective modal signal data; and inputting the reconstructed effective modal signal data into a sound-vibration collaborative recognition model, and outputting a structural state recognition result of the target special equipment through the sound-vibration collaborative recognition model. According to the method and the device, the problem of low sound vibration detection accuracy of a traditional sound vibration detection technology when mechanical impact, fluid disturbance, electromagnetic interference or multi-source background noise is accompanied in the environment where the detected equipment is located is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of acoustic vibration detection, and in particular to an anti-interference acoustic vibration detection method and device. BACKGROUND

[0002] Nowadays, by fusing a multi-channel high-sensitivity sensor array with a modal identification algorithm, through a multi-stage processing link of signal preprocessing, modal reconstruction, feature extraction and intelligent identification, fine identification and fault trend quantitative evaluation of the structure state can be realized under high-noise background or complex working conditions, which has good engineering adaptability and technical prospect.

[0003] However, in the traditional acoustic vibration detection technology, since a single-channel acquisition and static frequency domain analysis strategy are generally adopted, when mechanical impact, fluid disturbance, electromagnetic interference or multi-source background noise are accompanied in the environment where the device is located, the effective modal components in the acoustic vibration signal are often covered by random interference or structural coupling effect, resulting in serious modal aliasing, frequency drift or pseudo-response phenomenon, thereby significantly reducing the accuracy of subsequent acoustic vibration detection.

[0004] Therefore, an anti-interference acoustic vibration detection method and device are urgently needed. SUMMARY

[0005] The present application provides an anti-interference acoustic vibration detection method and device, which solves the problem of low accuracy of traditional acoustic vibration detection technology when mechanical impact, fluid disturbance, electromagnetic interference or multi-source background noise are accompanied in the environment where the device to be detected is located.

[0006] In a first aspect of the present application, an anti-interference acoustic vibration detection method is provided, which comprises: acquiring a target special device and a device type corresponding to the target special device; acquiring a multi-channel high-sensitivity sensor array combination corresponding to the device type in a preset inspection method database according to the device type, one device type corresponding to one multi-channel high-sensitivity sensor array combination; acquiring acoustic vibration signal data corresponding to the target special device based on the multi-channel high-sensitivity sensor array combination; applying an improved empirical mode decomposition algorithm to modal decomposition of the acoustic vibration signal data, eliminating pseudo-modal components and reconstructing effective modal signal data; inputting the reconstructed effective modal signal data into an acoustic vibration collaborative recognition model, and outputting a structure state recognition result of the target special device through the acoustic vibration collaborative recognition model.

[0007] Optionally, before the target special equipment and the equipment type corresponding to the target special equipment are acquired, the method further comprises: classifying the target special equipment according to a structure and a function attribute category of the special equipment, and classifying the special equipment into a plurality of equipment types, the equipment type corresponding to the target special equipment being any one of the plurality of equipment types, and the plurality of equipment types including pressure-bearing equipment, hoisting equipment, elevator equipment, aerial work equipment, pressure special environment equipment, rail transit equipment, and aerospace equipment.

[0008] Optionally, before the multi-channel high-sensitivity sensing array combination corresponding to the equipment type is acquired from the preset inspection method database according to the equipment type, the preset inspection method database needs to be constructed, and specifically comprises: constructing a correspondence between the equipment type and the multi-channel high-sensitivity sensing array combination based on different equipment types and according to a preset correspondence mode, the preset correspondence mode including a sensor type, a number of acquisition channels, a sensor spatial arrangement mode, inter-channel synchronous sampling accuracy, and a signal acquisition frequency band range configuration mode; and storing the correspondence in the preset inspection method database to construct the preset inspection method database.

[0009] Optionally, the multi-channel high-sensitivity sensing array combination includes one or more combinations of a miniature acceleration sensor array, a laser vibration sensing sensor array, an acoustic emission sensor array, a piezoelectric sensor array, a microwave vibration probe array, a magnetostrictive sensor array, a wideband capacitive microphone array, and a directional ultrasonic sensor array, and when the equipment type is pressure-bearing equipment, the acoustic vibration signal data corresponding to the target special equipment is acquired based on the multi-channel high-sensitivity sensing array combination, and specifically comprises: taking the acoustic emission sensor array, the piezoelectric sensor array, the miniature acceleration sensor array, and the directional ultrasonic sensor array as the multi-channel high-sensitivity sensing array combination; acquiring acoustic emission transient pulse data based on the acoustic emission sensor array, acquiring harmonic response signal data based on the piezoelectric sensor array, acquiring acceleration time-domain response data based on the miniature acceleration sensor array, and acquiring ultrasonic echo feature data based on the directional ultrasonic sensor array; and taking the acoustic emission transient pulse data, the harmonic response signal data, the acceleration time-domain response data, and the ultrasonic echo feature data as the acoustic vibration signal data.

[0010] Optionally, the improved empirical mode decomposition algorithm is applied to the sound and vibration signal data for modal decomposition, false modal components are removed, and valid modal signal data is reconstructed, specifically including: performing endpoint extension processing and white noise superposition processing on the sound and vibration signal data to enhance the separability of the modal components corresponding to the sound and vibration signal data; performing ensemble empirical mode decomposition operation on the enhanced sound and vibration signal data to obtain an initial intrinsic mode function component sequence; calculating the modal characteristic parameter values corresponding to the initial intrinsic mode function component sequence through the improved empirical mode decomposition algorithm, the modal characteristic parameter values including energy proportion value, frequency stability value, and instantaneous frequency fluctuation value; based on the modal characteristic parameter values, identifying and removing the corresponding false modal components in the sound and vibration signal data according to a preset false modal discrimination rule; and accumulating and reconstructing the remaining modal components after removal to generate valid modal signal data.

[0011] Optionally, the reconstructed valid modal signal data is input into a sound and vibration collaborative recognition model, and a structure state recognition result of the target special equipment is output through the sound and vibration collaborative recognition model, specifically including: obtaining historical sound and vibration signal data corresponding to the identified special equipment; applying the improved empirical mode decomposition algorithm to the historical sound and vibration signal data for modal decomposition, removing false modal components, and reconstructing historical valid modal signal data; obtaining multi-scale feature parameters based on the historical valid modal signal data, the multi-scale feature parameters including frequency distribution feature parameters, energy entropy feature parameters, envelope variation coefficient parameters, and transient response amplitude change rate parameters; obtaining equipment state labels corresponding to the multi-scale feature parameters, and constructing a training sample set based on the multi-scale feature parameters and the equipment state labels, the equipment state labels including structure integrity level labels, operating condition labels, defect type labels, numerical classification labels, and time sequence evolution labels; constructing a multi-layer neural network structure based on the training sample set, the neural network structure including feature extraction layers, cross-channel attention fusion layers, and state discrimination output layers corresponding to sound channels and vibration channels, respectively; training the neural network through a supervised learning method, and constructing the sound and vibration collaborative recognition model.

[0012] Optionally, the reconstructed effective modal signal data is input into the sound-vibration collaborative identification model, and a structure state identification result of the target special equipment is output through the sound-vibration collaborative identification model, specifically including: based on the effective modal signal data, extracting acoustic spectrum features and vibration response features through the sound-vibration collaborative identification model; weighting and fusing the acoustic spectrum features and the vibration response features in a cross-channel attention fusion structure, and generating a joint feature vector; according to the joint feature vector, calculating a target sound-vibration risk score value through the sound-vibration collaborative identification model; judging the size relationship between the target sound-vibration risk score value and a preset state threshold; if the target sound-vibration risk score value is greater than the preset state threshold, outputting the sound-vibration detection identification result as an abnormal sound-vibration detection identification result; if the target sound-vibration risk score value is less than or equal to the preset state threshold, outputting the sound-vibration detection identification result as a normal sound-vibration detection identification result.

[0013] Optionally, the defect type state label includes a cavity defect type state label and a delamination defect type state label, and after the reconstructed effective modal signal data is input into the sound-vibration collaborative identification model and a structure state identification result of the target special equipment is output through the sound-vibration collaborative identification model, the method further includes: obtaining first training sample data corresponding to the delamination defect type state label, and obtaining second training sample data corresponding to the delamination defect type state label; constructing a defect type discrimination sub-model based on the first training sample data and the second training sample data, and integrating the defect type discrimination sub-model into the sound-vibration collaborative identification model to form a multi-task identification structure; inputting the reconstructed effective modal signal data into the defect type discrimination sub-model, and outputting a defect type probability distribution corresponding to the target special equipment based on the defect type discrimination sub-model; and outputting a target defect type corresponding to the target special equipment through the defect type discrimination sub-model based on the defect type probability distribution.

[0014] In a second aspect of the present application, an anti-interference sound-vibration detection device is provided, which includes an acquisition module and a processing module, wherein, The acquisition module is configured to acquire a target special equipment and a device type corresponding to the target special equipment; based on the device type, acquire a multi-channel high-sensitivity sensor array combination corresponding to the device type in a preset inspection method database, one device type corresponding to one multi-channel high-sensitivity sensor array combination; and acquire sound-vibration signal data corresponding to the target special equipment based on the multi-channel high-sensitivity sensor array combination.

[0015] The processing module is configured to apply an improved empirical mode decomposition algorithm to the sound-vibration signal data for modal decomposition, eliminate pseudo-modal components, and reconstruct effective modal signal data; input the reconstructed effective modal signal data into the sound-vibration collaborative identification model, and output a structure state identification result of the target special equipment through the sound-vibration collaborative identification model.

[0016] In a third aspect of the present application, an electronic device is provided, comprising a processor, a memory, a user interface and a network interface, the memory is configured to store instructions, the user interface and the network interface are configured to communicate with other devices, and the processor is configured to execute the instructions stored in the memory to enable the electronic device to perform the method of any one of the above.

[0017] In a fourth aspect of the present application, a computer-readable storage medium is provided, which stores a computer program, and the computer program is configured to enable a processor to perform the method of any one of the above.

[0018] The one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. Obtain a target special equipment and a device type corresponding to the target special equipment; according to the device type, obtain a multi-channel high-sensitivity sensor array combination corresponding to the device type in a preset inspection method database; obtain a sound and vibration signal data corresponding to the target special equipment based on the multi-channel high-sensitivity sensor array combination; apply an improved empirical mode decomposition algorithm to the sound and vibration signal data for modal decomposition, eliminate false modal components and reconstruct effective modal signal data; input the reconstructed effective modal signal data into a sound and vibration collaborative recognition model, and output a structure state recognition result of the target special equipment through the sound and vibration collaborative recognition model, thereby improving the structure state recognition precision and robustness of different types of special equipment in a high-noise, multi-source interference environment, enhancing the early perception ability of complex structure defects, improving the adaptability, intelligent level and engineering practicability of the detection system, and realizing high credibility, rapid discrimination and intelligent risk warning of the structure health state of the special equipment.

[0019] 2. Based on different device types, a corresponding relationship between the device types and the multi-channel high-sensitivity sensor array combination is constructed in a preset corresponding manner, and the preset corresponding manner includes the configuration modes of sensor type, number of acquisition channels, sensor spatial layout, inter-channel synchronization sampling accuracy and signal acquisition frequency band range; the corresponding relationship is stored in the preset inspection method database to construct the preset inspection method database, thereby realizing adaptive calling of the optimal sensing configuration scheme for special equipment with different structural characteristics and working environments, ensuring that the collected sound and vibration signals have high separability, high signal-to-noise ratio and complete structural response coverage, providing a stable and reliable data basis for subsequent modal decomposition and state recognition, and effectively improving the universality, accuracy and deployment flexibility of the detection system in a multi-device scenario.

[0020] 3. The improved empirical mode decomposition algorithm is applied to the historical vibration signal data for modal decomposition, false modal components are removed, and historical effective modal signal data are reconstructed; multi-scale feature parameters are obtained based on the historical effective modal signal data; device state labels corresponding to the multi-scale feature parameters are obtained, and a training sample set is constructed based on the multi-scale feature parameters and the device state labels; a multi-layer neural network structure is constructed based on the training sample set, the neural network structure includes feature extraction layers, cross-channel attention fusion layers, and state discrimination output layers corresponding to sound channels and vibration channels respectively; the neural network is trained through a supervised learning method, and a sound-vibration collaborative recognition model is constructed, thereby realizing a high-dimensional supervised learning mechanism that fuses historical structural response features and state labels, enabling the constructed sound-vibration collaborative recognition model to have state discrimination ability and generalization ability under cross-modal, multi-working-condition, and heterogeneous structures, effectively improving the recognition depth of the model on complex sound-vibration coupling features and the classification accuracy of abnormal states, and providing a transferable and expandable intelligent structural health monitoring solution for special equipment. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 is a flowchart of an anti-interference sound-vibration detection method provided by an embodiment of the present application; Figure 2 is a module schematic diagram of an anti-interference sound-vibration detection device provided by an embodiment of the present application; Figure 3 is a structural schematic diagram of an electronic device provided by an embodiment of the present application.

[0022] Legend of reference signs: 21, acquisition module; 22, processing module; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. DETAILED DESCRIPTION

[0023] In order for those skilled in the art to better understand the technical solutions in the specification, the technical solutions in the specification will be clearly and completely described below in combination with the drawings in the embodiments of the specification. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments.

[0024] The terms used in the following embodiments of the present application are only for the purpose of describing the specific embodiments, and are not intended to be limiting to the present application. As used in the specification of the present application, the singular expression "one", "a", "said", "the above", "the", and "this" are intended to also include the plural expression, unless there is clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application means and includes any or all possible combinations of one or more listed items.

[0025] Hereinafter, the terms "first", "second" are only used for descriptive purposes, and cannot be understood as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, the meaning of "multiple" is two or more, unless otherwise specified.

[0026] In order for those skilled in the art to better understand the technical solutions of the present application, the present application will be further described in detail below with reference to the drawings.

[0027] Please refer to Figure 1 which shows a flowchart of an anti-interference acoustic vibration detection method provided by an embodiment of the present application. The flowchart mainly includes the following steps: S101 to S105.

[0028] Step S101, obtaining a target special equipment and a device type corresponding to the target special equipment.

[0029] Specifically, the target special equipment is classified according to the structure and functional attribute category of the special equipment, and the special equipment is divided into multiple device types with typical structural characteristics and risk scene characteristics. The device type corresponding to the target special equipment is any one of the multiple device types, and the device type includes pressure-bearing equipment, hoisting equipment, elevator equipment, high-altitude operation equipment, pressure special environment equipment, rail transit equipment and aerospace equipment. For each device type, its typical use scene and specific equipment are as follows: Pressure-bearing equipment is mainly used for storing, conveying or converting gas or liquid medium with high pressure characteristics, and is widely used in petroleum chemical industry, power, metallurgy, heat and pharmaceutical industry, etc. Its operation process has major safety hazards such as explosion, leakage and overpressure. Typical equipment includes fixed pressure vessels (such as reaction kettles, gas storage tanks, liquid ammonia tanks), mobile pressure vessels (such as liquefied gas transport tank trucks), boilers (such as steam boilers, hot water boilers, organic heat carrier boilers), gas cylinders, heat exchangers, etc.

[0030] Hoisting equipment is used to realize the lifting, carrying and hoisting of materials or components in vertical or horizontal direction, and is widely used in construction sites, ports, warehouses, logistics and manufacturing assembly scenes. Typical equipment includes bridge cranes, gantry cranes, tower cranes, portal cranes, mobile cranes, cable cranes, electric hoists, elevators and their attached cantilever structures and hoisting mechanisms, etc.

[0031] Elevator-type equipment is mainly used for vertical lifting or inclined conveying of personnel or goods within a building or facility, with high operation frequency and wide audience range, and is often used in residential buildings, office buildings, airports, shopping malls, rail station buildings, etc. Typical equipment includes passenger elevators, cargo elevators, escalators, moving sidewalks, fire elevators, and medical elevators, etc. Key structures involve traction systems, door systems, guide rail systems, and control and drive systems.

[0032] Aerial work equipment is used to realize the lifting, hovering and attitude control of workers or equipment in high places, and is often used in power maintenance, building decoration, communication base station maintenance and ship repair scenes. Typical equipment includes scissor lift platforms, mast lift platforms, telescopic arm climbing platforms, aerial work vehicles, building baskets, and crawler lifts, etc. The structural design needs to consider platform stability and anti-overturning ability.

[0033] Pressure special environment equipment refers to structural systems operating in extreme environments such as high temperature, high pressure, high radiation, deep water, and high corrosion, with high failure risk and high environmental coupling. It is often used in nuclear power plants, offshore platforms, deep well drilling, and aircraft engine test systems. Typical equipment includes reactor pressure shells, main steam pipes, pressurizers, subsea manifolds, sealed cabin bodies, deep water pressure test chambers, and high temperature cracking systems.

[0034] Rail transit equipment refers to transportation equipment and its attached structures operating on rail infrastructure, with high-speed operation, high vibration and impact, and long-term service requirements, widely used in urban rail, subway, high-speed rail, etc. Typical equipment includes traction motor systems, bogies, coupler buffer devices, brake systems, switch actuators, pantograph devices, and vibration reduction components, etc.

[0035] Aerospace equipment is used for aircraft body, propulsion system, attitude control system and cabin structure, with service environment involving high altitude, vacuum, high speed and strong thermal vibration coupling field, commonly used in civil aviation, military, space exploration and satellite launch platform. Typical equipment includes aircraft landing gear, wing main beam, tail structure, gas turbine blade, engine casing, satellite support cabin, hypersonic aerodynamic cabin and aircraft connection frame, etc.

[0036] Through the above classification and analysis, the structural attribute identification and use scene positioning of any target special equipment can be realized, thereby providing basic input for subsequent matching of sensor array configuration and acoustic vibration detection strategy.

[0037] In step S102, according to the type of equipment, a multi-channel high-sensitivity sensor array combination corresponding to the type of equipment is obtained from a preset inspection method database.

[0038] Specifically, by identifying the equipment type to which the target special equipment belongs, a sensor layout scheme corresponding to the equipment type is found and called in the constructed preset inspection method database, so that a multi-channel high-sensitivity sensor array combination matched with the target special equipment is determined for subsequent acoustic vibration signal data acquisition operations.

[0039] In a possible implementation, the step S102 further includes: classifying the target special equipment according to structural and functional attribute categories of the special equipment, classifying the special equipment into a plurality of equipment types, and the equipment type corresponding to the target special equipment being any one of the plurality of equipment types, the plurality of equipment types including pressure-bearing equipment, hoisting equipment, elevator equipment, high-altitude operation equipment, pressure special environment equipment, rail transit equipment, and aerospace equipment.

[0040] Specifically, based on different equipment types, a preset corresponding relationship between the equipment types and the multi-channel high-sensitivity sensor array combination is constructed in a preset corresponding manner, the preset corresponding manner including configurations of sensor types, acquisition channel numbers, sensor spatial layout manners, inter-channel synchronous sampling accuracies, and signal acquisition frequency band ranges. The sensor types are selected according to detection target parts and defect sensitivities, and include acoustic emission sensors, miniature acceleration sensors, piezoelectric sensors, laser vibration measurement sensors, directional ultrasonic sensors, etc.; the acquisition channel numbers are set according to equipment structural sizes and coverage accuracy requirements, to ensure that signal capture has distribution and redundancy; the sensor spatial layout manner is arranged according to equipment surface geometric structures and key stress areas, and includes ring distribution, longitudinal layout, matrix array, etc.; the inter-channel synchronous sampling accuracy is used to ensure the consistency of signals in the time domain and avoid data distortion caused by phase errors; and the signal acquisition frequency band range should cover the response frequency domain of typical defects of the equipment, to ensure effective capture of different modal signals. After construction, the above equipment types and their corresponding sensor array configuration schemes are indexed and uniformly stored in the preset inspection method database, to form an acoustic vibration detection configuration knowledge base for different equipment types, supporting subsequent rapid matching and calling.

[0041] In step S103, acoustic vibration signal data corresponding to the target special equipment is acquired based on the multi-channel high-sensitivity sensor array combination.

[0042] Specifically, according to the multi-channel high-sensitivity sensor array combination corresponding to the target special equipment determined in step S102, layout and calibration operations of various sensors at key parts of the structure of the target special equipment are completed, and acoustic signals and structural vibration signals generated by the equipment under actual operating conditions or excitation response conditions are collected and recorded in parallel, so as to acquire original acoustic vibration signal data covering multiple scales, multiple frequency domains, and multiple channels.

[0043] The acquisition process includes: according to the device structure geometry and the typical defect propagation path, fixing and installing each sensor unit at the structure weld, the stress concentration part, the vibration node area, the sealed interface, the pressure shell area and other acoustic vibration sensitive positions; through the high-precision synchronous sampling system, all sensor channels are uniformly controlled to ensure that the obtained acoustic signals and vibration signals are strictly aligned in the time dimension, and the phase offset and feature aliasing caused by multi-channel data asynchronization are avoided; during the acquisition period, the dynamic behavior of the device under natural operation, loading response or specific detection conditions is monitored throughout the cycle, and various types of data such as acoustic emission signals, acceleration response, harmonic amplitude-frequency characteristics and ultrasonic echo signals are digitized and collected, and the acoustic vibration signal data set in a unified format is output.

[0044] In addition, in order to improve the effectiveness and anti-interference ability of the signal, the collected data can be preliminarily filtered and stabilized by combining multi-channel filtering algorithm, adaptive gain adjustment technology and environmental noise identification mechanism, and the non-structural interference signals introduced by external disturbance, base coupling or electromagnetic interference are removed, so as to form the original acoustic vibration signal data with high time-frequency accuracy, structural consistency and identification value, and provide basic input for subsequent modal decomposition and structural state discrimination.

[0045] In one possible implementation, the multi-channel high-sensitivity sensor array combination includes one or more combinations of a micro-acceleration sensor array, a laser vibration sensor array, an acoustic emission sensor array, a piezoelectric sensor array, a microwave vibration probe array, a magnetostrictive sensor array, a wideband capacitive microphone array and a directional ultrasonic sensor array. When the device type is a pressure-bearing device, step S103 further includes: using the acoustic emission sensor array, the piezoelectric sensor array, the micro-acceleration sensor array and the directional ultrasonic sensor array as the multi-channel high-sensitivity sensor array combination; acquiring acoustic emission transient pulse data based on the acoustic emission sensor array, harmonic response signal data based on the piezoelectric sensor array, acceleration time domain response data based on the micro-acceleration sensor array, and ultrasonic echo feature data based on the directional ultrasonic sensor array; and using the acoustic emission transient pulse data, the harmonic response signal data, the acceleration time domain response data and the ultrasonic echo feature data as the acoustic vibration signal data.

[0046] Specifically, for example, when the device type is a pressure-bearing device, due to the characteristics of strong airtightness, high bearing pressure, many welded structures, and complex service environment, common failure forms include shell cracks, weld leaks, fatigue damage, and interface delamination, etc. Therefore, an array of acoustic emission sensors, an array of piezoelectric sensors, an array of miniature acceleration sensors, and an array of directional ultrasonic sensors are combined as a multi-channel high-sensitivity sensor array. At this time, acoustic emission transient pulse data is obtained based on the array of acoustic emission sensors, which includes but is not limited to crack propagation pulse signals, leakage release high-frequency signals, material friction shock wave signals, and acoustic source positioning time difference signals; harmonic response signal data is obtained based on the array of piezoelectric sensors, which includes but is not limited to local modal response amplitude, frequency spectrum amplitude-frequency distribution, frequency drift rate, and nonlinear harmonic intensity; acceleration time domain response data is obtained based on the array of miniature acceleration sensors, which includes instantaneous acceleration amplitude, impact response peak, structural vibration frequency, and oscillation attenuation coefficient; ultrasonic echo feature data is obtained based on the array of directional ultrasonic sensors, which includes sound wave propagation delay time, interface reflection amplitude, echo phase shift, and sound path abnormality ratio.

[0047] For example, when the device type is an elevator device, due to the fact that an elevator system usually includes a traction mechanism, a guide rail system, a door machine system, and a car structure, problems such as guide rail abnormal noise, steel wire rope broken strands, brake failure, and damping element loosening are prone to occur during operation, therefore, the array of miniature acceleration sensors, the array of piezoelectric sensors, and the array of acoustic emission sensors are preferably jointly arranged to collect guide rail vibration response, load variation harmonic, and abnormal friction sound source information, thereby improving the real-time recognition ability of sudden structural abnormalities.

[0048] For example, when the device type is a high-altitude operation device, due to the fact that the platform lifting process involves complex scissors structures, hydraulic supports, and telescopic arm mechanisms, there are risks of support instability, joint loosening, and loading mutation, therefore, the array of laser vibration sensors and the array of miniature acceleration sensors are suitable for high-frequency dynamic monitoring to obtain platform vibration response and structural stiffness change characteristics.

[0049] For example, when the device type is an aerospace device, due to the fact that its structure is long-term in a high-vibration, high-thermal shock, and complex aerodynamic load environment, fatigue cracks, connection loosening, or high-frequency vibration resonance are prone to occur, therefore, a high-response, multi-dimensional sensing system is constructed by using the array of piezoelectric sensors, the array of acoustic emission sensors, and the array of microwave vibration probes, to realize multi-modal joint monitoring of the wing spar, the engine shell, and the cabin frame, thereby ensuring the stability and safety of the flight structure state.

[0050] Step S104, applying the improved empirical mode decomposition algorithm to the acoustic vibration signal data for modal decomposition, eliminating the pseudo-modal components and reconstructing the effective modal signal data.

[0051] Specifically, after preprocessing the acquired original acoustic vibration signal data, the improved ensemble empirical mode decomposition algorithm is used to extract components from the signal, obtain a group of intrinsic mode function components with local characteristics, and identify and eliminate pseudo-modal components that are not true responses by quantitatively analyzing the energy proportion value, frequency stability value and instantaneous frequency fluctuation value of each component; the remaining true modal components are reconstructed to obtain effective modal signal data representing the actual acoustic vibration response characteristics of the equipment, which are used as input signals for subsequent state identification analysis.

[0052] In one possible implementation, step S104 further includes: performing endpoint extension processing and white noise superposition processing on the acoustic vibration signal data to enhance the separability of the modal components corresponding to the acoustic vibration signal data; performing ensemble empirical mode decomposition operation on the enhanced acoustic vibration signal data to obtain an initial intrinsic mode function component sequence; calculating modal characteristic parameter values corresponding to the initial intrinsic mode function component sequence by the improved empirical mode decomposition algorithm, the modal characteristic parameter values including energy proportion value, frequency stability value and instantaneous frequency fluctuation value; identifying and eliminating corresponding pseudo-modal components in the acoustic vibration signal data based on the modal characteristic parameter values according to a preset pseudo-modal discrimination rule; and accumulating and reconstructing the remaining modal components after elimination to generate effective modal signal data.

[0053] Specifically, the endpoint extension processing and white noise superposition processing are performed on the acoustic vibration signal data to enhance the separability of the modal components corresponding to the acoustic vibration signal data: in order to prevent the end effect in the empirical mode decomposition process from causing modal aliasing or boundary distortion, the original signal is first linearly extrapolated at both ends, and then controlled-amplitude Gaussian white noise is superimposed to enhance the separability of modal separation:

[0054] wherein, represents the acoustic vibration signal data of the i th sensing channel after the endpoint extension processing and the white noise superposition processing, which is used to improve the separation ability of the modal components in the boundary region of the signal and suppress the modal aliasing caused by the boundary effect, if the system includes N sensing channels, then , represents the original acoustic vibration signal data value of the i th sensing channel at the initial time t 0, represents the original acoustic vibration signal data value of the i th sensing channel at the time t 1, represents the original acoustic vibration signal data value of the i th sensing channel at the time t 2, represents the original acoustic vibration signal data value of the i th sensing channel at the time t 3, represents the original acoustic vibration signal data value of the i th sensing channel at the time t 4, represents the original acoustic vibration signal data value of the i th sensing channel at the time t 5, represents the original acoustic vibration signal data value of the i th sensing channel at the time t 6, ​denotes the original vibration signal data value of the th sensing channel at the time instant corresponding to the original vibration signal data value, denotes the original vibration signal data value of the th sensing channel at the terminal time instant , which is used to construct the linear extrapolation extension term of the signal terminal, denotes the continuous time variable or discrete sampling index, which is used to depict the variation process of the signal in the time domain, denotes the total time length, i.e., the sampling point number or the maximum time index value of the original vibration signal on each channel. Then, a noisy disturbance is added:

[0055] wherein, denotes the signal data of the th channel after enhancement processing, which will be used as the input signal of the ensemble decomposition stage for subsequent extraction of stable modal functions, denotes the white noise amplitude control coefficient of the th channel, which is usually a positive real number and is used to adjust the noise energy proportion, denotes the standard Gaussian white noise data of the th channel at the time instant , which satisfies .

[0056] The enhanced vibration signal data is subjected to an ensemble empirical mode decomposition operation to obtain an initial intrinsic modal function component sequence: the enhanced signal is subjected to multiple EMD operations with random noise disturbance to form an ensemble modal representation:

[0057] wherein, denotes the EMD decomposition result corresponding to the enhanced signal of the th sensing channel after the th time addition of independent white noise disturbance, denotes the number of intrinsic modal function (IMF) components decomposed from the enhanced signal of the th channel in the th decomposition, which is usually automatically determined by the signal structure complexity, denotes the th intrinsic modal function component decomposed from the th sensing channel under the th disturbance, which is an intrinsic vibration mode satisfying certain time domain symmetry and unimodality requirements, denotes the th intrinsic modal function component decomposed from the th The residual signal in the secondary decomposition contains the trend item or high-order low-frequency drift item that fails to be effectively extracted, and is usually the last remaining non-decomposable part.

[0058] The modal characteristic parameter values corresponding to the initial intrinsic mode function component sequence are calculated by the improved empirical mode decomposition algorithm, including energy proportion value, frequency stability value and instantaneous frequency fluctuation value. The energy proportion value is used to measure the energy contribution of the mode, and the mode with low energy proportion is usually noise or a pseudo mode, and the calculation method is as follows:

[0059] Wherein, represents the energy proportion value of the mth modal component of the nth channel, represents the total time length or sampling sequence length of the signal time domain. The calculation method of the frequency stability value is to extract the instantaneous frequency first: Wherein, represents the instantaneous frequency of the mth modal component of the nth channel,

[0060] represents the Hilbert transform, represents the first derivative of the time variable , reflecting the phase change rate. Then calculate the coefficient of variation: Wherein, represents the frequency stability value of the mth mode of the nth channel, represents the time average value of the instantaneous frequency, used for normalizing the fluctuation amplitude. The second-order perturbation size of the frequency sequence is used to measure the high-order fluctuation intensity, which indicates the discontinuity or distortion characteristics of the mode, and the calculation method is as follows:

[0061] Wherein, represents the instantaneous frequency fluctuation value corresponding to the mth intrinsic mode function in the nth sensing channel, used to measure the intensity of the frequency change of the modal component in the whole analysis time interval [0, T]. Wherein,

[0062] represents the instantaneous frequency fluctuation value corresponding to the mth intrinsic mode function in the nth sensing channel, used to measure the intensity of the frequency change of the modal component in the whole analysis time interval [0, T]. Based on the modal characteristic parameter values, the corresponding pseudo modal components in the acoustic vibration signal data are identified and removed according to the preset pseudo modal discrimination rule: a pseudo modal criterion set is constructed by combining the three types of characteristic parameters:

[0063] Based on the modal characteristic parameter values, the corresponding pseudo modal components in the acoustic vibration signal data are identified and removed according to the preset pseudo modal discrimination rule: a pseudo modal criterion set is constructed by combining the three types of characteristic parameters: ​​​​​

[0064] wherein, represents the th eigenmode function component in the th sensing channel, represents the discriminant threshold of energy proportion value, used to measure the effectiveness of the proportion of a modal component in the overall signal energy, when the energy proportion value of the modal component is lower than , it is determined that its energy contribution is insufficient, which is a disturbing or numerically false component, represents the discriminant threshold of frequency stability value, used to identify whether the modal frequency change is stable, when the relative fluctuation degree of modal frequency is higher than , it indicates that the modal has frequency drift or non-stationary oscillation, which may be a pseudo-modal or greatly affected by background disturbance, represents the discriminant threshold of instantaneous frequency fluctuation value, used to evaluate the degree of instantaneous frequency fluctuation of the modal component in the time domain, when the instantaneous frequency fluctuation value is higher

[0065] than , it indicates that the modal has high-frequency oscillation or uncontrollable disturbance in the local time window, which can be regarded as a non-structural signal component without physical interpretability.

[0066] wherein, is the effective modal signal data reconstructed by all effective eigenmode function components in the th sensing channel, is the index set of effective modal components, The setting mode of includes: energy proportion value threshold screening: only the modal components satisfying are retained, and the interference signals with lower energy contribution are removed; only the modal components satisfying are retained, and the components with high frequency drift or non-stationary behavior are removed; only the modal components satisfying are retained, and the pseudo-modal with dramatic frequency change and poor physical stability is removed. Based on the above three criteria, the joint discriminant rule is set as follows: .

[0067] In step S105, the reconstructed effective modal signal data is input into the sound-vibration cooperative identification model, and the structure state recognition result of the target special equipment is output through the sound-vibration cooperative identification model.

[0068] Specifically, based on the multi-channel effective modal signal data obtained in step S104, the acoustic signal feature channel and the vibration signal feature channel are constructed respectively, the characteristics of different physical sources are independently extracted, and cross-channel fusion learning is performed in the acoustic-vibration collaborative recognition model to realize comprehensive discrimination of the target special equipment structure state. In this step, the input effective modal signal data is a structured vector form after merging and normalization after feature engineering (such as STFT, envelope analysis, wavelet energy spectrum, empirical entropy, skewness, HHT frequency distribution, etc.) extraction on multiple channels, and the essence is:

[0069] wherein, is a high-dimensional feature vector finally used for input to the acoustic-vibration collaborative recognition model, represents a feature extraction operation function.

[0070] In one possible implementation, step S105 further includes: obtaining historical acoustic-vibration signal data corresponding to the identified special equipment; applying the improved empirical mode decomposition algorithm to the historical acoustic-vibration signal data for modal decomposition, eliminating the pseudo-modal components and reconstructing the historical effective modal signal data; obtaining multi-scale feature parameters based on the historical effective modal signal data, the multi-scale feature parameters including frequency distribution feature parameters, energy entropy feature parameters, envelope variation coefficient parameters, and transient response amplitude change rate parameters; obtaining equipment state labels corresponding to the multi-scale feature parameters, and constructing a training sample set based on the multi-scale feature parameters and the equipment state labels, the equipment state labels including structure integrity level labels, operating condition labels, defect type labels, numerical classification labels, and time sequence evolution labels; constructing a multi-layer neural network structure based on the training sample set, the neural network structure including feature extraction layers corresponding to acoustic channels and vibration channels respectively, cross-channel attention fusion layers, and state discrimination output layers; training the neural network through a supervised learning method, and constructing an acoustic-vibration collaborative recognition model.

[0071] Specifically, the historical acoustic-vibration signal data corresponding to the identified special equipment is obtained, and the improved empirical mode decomposition algorithm consistent with the current equipment is applied to the historical acoustic-vibration signal data to eliminate the pseudo-modal components and reconstruct the historical effective modal signal data. Then, multi-scale feature parameters are extracted based on the reconstructed historical effective modal signal data, including: Statistical features are extracted:

[0072]

[0073] ​

[0074] where, is the average frequency, is the modal frequency distribution sequence is the sample length, is the frequency standard deviation, is the frequency skewness coefficient, is the index variable of the frequency distribution sequence, taking the value range . Then, the modal energy entropy is calculated according to the frequency energy distribution probability

[0075] where, is the modal energy entropy. Let the modal signal envelope be , and its coefficient of variation is:

[0076] where, and are the mean and standard deviation of the envelope, respectively. Let the signal instantaneous amplitude be , then the transient change rate is:

[0077] where, ​is the transient change rate. The above features constitute a high-dimensional feature vector, and are paired with device state labels obtained by artificial labeling or sensor recording to form a training sample set, including: structural integrity level labels, which represent the current health status of the target special equipment in terms of structural mechanical performance and integrity, reflecting whether there are structural abnormalities such as cracks, corrosion, disassembly, deformation, and the severity thereof. The label is usually in a multi-level classification manner, such as: “intact”, “slight damage”, “moderate damage”, “severe damage”, or represented by numerical levels as 0-3 levels, corresponding to different levels of residual structural strength or lower limit of safety factor; operating condition labels, which represent the actual operating state of the equipment at the time of data collection, including load state, dynamic response state or environmental boundary conditions, such as: “steady-state operation”, “high-load operation”, “impact excitation”, “frequency resonance”, “unloading recovery”, etc. This label helps to distinguish whether the structural abnormalities are caused by external disturbances or internal defects; defect type labels, which represent the known typical defect modes or fault positions of the equipment at the time of data collection, with high interpretability and controllability. The label includes, for example: “weld crack”, “loose connection”, “support instability”, “seal leakage”, “support fracture”, “cavity defect”, etc., which can be used as the target output dimension of the model anomaly recognition result, for subsequent fault location and maintenance response refinement; numerical classification labels, which represent the quantitative risk level or abnormal probability level of the structural state, suitable for regression output or continuous state estimation tasks. The label can be set as a continuous risk score value between 0 and 1, or an integer level such as 0 (safe), 1 (suspicious), 2 (warning), 3 (dangerous), and can be used in combination with a risk response model of the monitoring index for fine modeling; time evolution labels, which represent the stage labels of the device structural state in the process of time evolution, especially suitable for state tracking tasks in the context of fatigue degradation, aging failure or multi-cycle loading. The label includes, for example: “initial state”, “stable period”, “crack initiation period”, “damage propagation period”, “failure critical period”, etc. With time step information, the model can be trained to identify the structural degradation trend of the equipment, achieving predictive diagnosis and early warning. Then, a multi-layer neural network structure is constructed, including: a feature extraction layer: convolutional encoders are used to extract local features in the sound and vibration channels; an attention fusion layer: a cross-channel attention mechanism is introduced to weight and integrate the sound and vibration features; a state discrimination layer: a fully connected structure is used to output specific structural state categories or scores combined with Softmax or Sigmoid.

[0078] Finally, the model is trained based on the above training sample set in a supervised learning manner, minimizing the loss function:

[0079] wherein, represents the loss function, represents the true label, is a model prediction value, is a regularization coefficient, is a set of network parameters.

[0080] In one possible implementation, step S105 further includes: based on the effective modal signal data, extracting acoustic spectrum features and vibration response features by the acoustic-vibration collaborative identification model; weighting and fusing the acoustic spectrum features and the vibration response features in a cross-channel attention fusion structure, and generating a joint feature vector; according to the joint feature vector, calculating a target acoustic-vibration risk score value by the acoustic-vibration collaborative identification model; judging a size relationship between the target acoustic-vibration risk score value and a preset state threshold value; if the target acoustic-vibration risk score value is greater than the preset state threshold value, outputting the acoustic-vibration detection and identification result as an abnormal acoustic-vibration detection and identification result; if the target acoustic-vibration risk score value is less than or equal to the preset state threshold value, outputting the acoustic-vibration detection and identification result as a normal acoustic-vibration detection and identification result.

[0081] Specifically, based on the reconstructed effective modal signal data, acoustic spectrum feature vectors and vibration response feature vectors are extracted respectively. The extracted features include but are not limited to: power spectral density, multi-scale energy entropy, frequency domain principal component coefficient, envelope modulation statistics, instantaneous modal amplitude, and other multi-dimensional signal statistics.

[0082] By introducing an attention mechanism, acoustic channel weights and vibration channel weights are calculated, which respectively represent the degree of attention to acoustic and vibration channel features in dimensions In the channel fusion layer, a weighted joint feature vector is generated, and its expression is:

[0083] This operation dynamically adjusts the weights of each dimension to enhance the structural state related information and suppress the redundant modal interference. The joint feature vector is input into the structural state score layer, and a set of trained projection parameters and a bias term are used to calculate the target acoustic-vibration risk score value :

[0084] This score value represents the risk intensity of the current special equipment structural state, and is a quantitative index of structural failure trend after integrating all modal features. A state judgment threshold value is set, and the score value and the threshold value are classified according to their size relationship: if , an abnormal acoustic-vibration detection and identification result is output; if , a normal acoustic-vibration detection and identification result is output.

[0085] In a possible implementation, step S105 further includes: obtaining first training sample data corresponding to the hierarchical defect type state label, and obtaining second training sample data corresponding to the hierarchical defect type state label; constructing a defect type discrimination sub-model based on the first training sample data and the second training sample data, and integrating the defect type discrimination sub-model into the sound-vibration collaborative recognition model to form a multi-task recognition structure; inputting the reconstructed effective modal signal data into the defect type discrimination sub-model, and outputting a defect type probability distribution corresponding to the target special equipment based on the defect type discrimination sub-model; and outputting, by the defect type discrimination sub-model, a target defect type corresponding to the target special equipment based on the defect type probability distribution.

[0086] Specifically, let the defect feature vector dimension be , the feature dimension of the weighted joint feature vector is consistent, the number of cavity defect samples is , and the number of hierarchical defect samples is , wherein the common features in the cavity defect samples include: envelope waveform discontinuity increase, low-frequency energy abnormal aggregation, high-amplitude pulse concentrated distribution, etc.; the common features in the hierarchical defect samples include: specific direction echo attenuation, ultrasonic response periodicity destruction, modal frequency stability decline, etc. Then the cavity sample feature matrix is:

[0087] , wherein represents the feature matrix of the cavity defect sample, represents the label matrix corresponding to the cavity defect sample. The hierarchical sample feature matrix is:

[0088] , wherein represents the feature matrix of the hierarchical defect sample, represents the label matrix of the hierarchical defect sample. The output of the defect type discrimination sub-model is: , the model weight and the bias are defined, and the linear classifier output is:

[0089] , wherein is the linear classifier output, is the high-dimensional feature vector finally used for input into the sound-vibration collaborative recognition model. The output is subjected to Softmax normalization to obtain the defect type probability distribution:

[0090]

[0091]

[0092] wherein, is the probability that the current input sample belongs to the "cavity defect", represents the original discriminant score value corresponding to the "cavity defect" category after the weight mapping and bias translation of the high-dimensional feature vector by the linear classifier for the final input into the sound-vibration collaborative identification model, is the probability that the current input sample belongs to the "delamination defect", represents the original discriminant score value corresponding to the "delamination defect" category after the weight mapping and bias translation of the high-dimensional feature vector by the linear classifier for the final input into the sound-vibration collaborative identification model, represents the defect type probability distribution vector.

[0093] The defect type judgment rule is set as:

[0094] represents the defect type judgment rule, then when , the defect type discriminant sub-model outputs the "cavity defect type"; and when , the defect type discriminant sub-model outputs the "delamination defect type".

[0095] The present application realizes the full-process linkage identification from data-driven feature representation, cross-domain information fusion, to structure risk quantification and abnormal state classification, has strong generalization ability and real-time decision-making ability, and is suitable for high-interference and high-complexity special equipment structure state evaluation scenes.

[0096] Please refer to Figure 2 , which shows a module schematic diagram of an anti-interference sound-vibration detection device provided by an embodiment of the present application, and the device comprises an acquisition module 21 and a processing module 22, wherein, The acquisition module 21 is configured to acquire a target special equipment and a device type corresponding to the target special equipment; according to the device type, acquire a multi-channel high-sensitivity sensor array combination corresponding to the device type in a preset inspection method database, one device type corresponds to one multi-channel high-sensitivity sensor array combination; and acquire sound-vibration signal data corresponding to the target special equipment based on the multi-channel high-sensitivity sensor array combination.

[0097] The processing module 22 is configured to apply the improved empirical mode decomposition algorithm to the acoustic vibration signal data to perform modal decomposition, remove false modal components, and reconstruct valid modal signal data; input the reconstructed valid modal signal data into the acoustic vibration collaborative identification model, and output the structural state identification result of the target special equipment through the acoustic vibration collaborative identification model.

[0098] In a possible implementation, the acquisition module 21 is configured to, before acquiring the target special equipment and the device type corresponding to the target special equipment, the method further includes: classifying the target special equipment according to the structure and function attribute category of the special equipment, and classifying the special equipment into a plurality of device types, the device type corresponding to the target special equipment being any one of the plurality of device types, and the plurality of device types including pressure-bearing devices, hoisting devices, elevator devices, high-altitude operation devices, pressure special environment devices, rail transit devices, and aerospace devices.

[0099] In a possible implementation, the acquisition module 21 is configured to, before acquiring, according to the device type, the multi-channel high-sensitivity sensor array combination corresponding to the device type in the preset inspection method database, construct the preset inspection method database, specifically including: constructing a correspondence between the device type and the multi-channel high-sensitivity sensor array combination based on different device types and according to a preset correspondence mode, the preset correspondence mode including the configuration mode of the sensor type, the number of acquisition channels, the sensor spatial arrangement mode, the inter-channel synchronous sampling accuracy, and the signal acquisition frequency band range; and storing the correspondence in the preset inspection method database to construct the preset inspection method database.

[0100] In a possible implementation, the multi-channel high-sensitivity sensor array combination includes one or more combinations of a micro-acceleration sensor array, a laser vibration measurement sensor array, an acoustic emission sensor array, a piezoelectric sensor array, a microwave vibration probe array, a magnetostrictive sensor array, a wideband capacitive microphone array, and a directional ultrasonic sensor array. When the device type is a pressure-bearing device, the acquisition module 21 is configured to acquire the acoustic vibration signal data corresponding to the target special equipment based on the multi-channel high-sensitivity sensor array combination, specifically including: taking the acoustic emission sensor array, the piezoelectric sensor array, the micro-acceleration sensor array, and the directional ultrasonic sensor array as the multi-channel high-sensitivity sensor array combination; acquiring acoustic emission transient pulse data based on the acoustic emission sensor array, harmonic response signal data based on the piezoelectric sensor array, acceleration time-domain response data based on the micro-acceleration sensor array, and ultrasonic echo feature data based on the directional ultrasonic sensor array; and taking the acoustic emission transient pulse data, the harmonic response signal data, the acceleration time-domain response data, and the ultrasonic echo feature data as the acoustic vibration signal data.

[0101] In a possible implementation, the processing module 22 is configured to apply the improved empirical mode decomposition algorithm to the acoustic vibration signal data to perform modal decomposition, eliminate pseudo-modal components, and reconstruct valid modal signal data, specifically including: performing endpoint extension processing and white noise superposition processing on the acoustic vibration signal data to enhance the separability of the modal components corresponding to the acoustic vibration signal data; performing ensemble empirical mode decomposition operation on the enhanced acoustic vibration signal data to obtain an initial intrinsic mode function component sequence; calculating modal characteristic parameter values corresponding to the initial intrinsic mode function component sequence by the improved empirical mode decomposition algorithm, the modal characteristic parameter values including energy proportion value, frequency stability value, and instantaneous frequency fluctuation value; identifying and eliminating corresponding pseudo-modal components in the acoustic vibration signal data based on the modal characteristic parameter values and according to a preset pseudo-modal discrimination rule; and accumulating and reconstructing the remaining modal components after elimination to generate valid modal signal data.

[0102] In a possible implementation, the processing module 22 is configured to input the reconstructed valid modal signal data into the acoustic vibration collaborative recognition model, and output the structural state recognition result of the target special equipment by the acoustic vibration collaborative recognition model, specifically including: obtaining historical acoustic vibration signal data corresponding to the identified special equipment; applying the improved empirical mode decomposition algorithm to the historical acoustic vibration signal data to perform modal decomposition, eliminate pseudo-modal components, and reconstruct historical valid modal signal data; obtaining multi-scale feature parameters based on the historical valid modal signal data, the multi-scale feature parameters including frequency distribution feature parameters, energy entropy feature parameters, envelope variation coefficient parameters, and transient response amplitude change rate parameters; obtaining equipment state labels corresponding to the multi-scale feature parameters, and constructing a training sample set based on the multi-scale feature parameters and the equipment state labels, the equipment state labels including structural integrity level labels, operating condition labels, defect type labels, numerical classification labels, and time sequence evolution labels; constructing a multi-layer neural network structure based on the training sample set, the neural network structure including feature extraction layers, cross-channel attention fusion layers, and state discrimination output layers corresponding to acoustic channels and vibration channels, respectively; training the neural network in a supervised learning manner, and constructing the acoustic vibration collaborative recognition model.

[0103] In a possible implementation, the processing module 22 is configured to input the reconstructed effective modal signal data into the acoustic-vibration collaborative identification model, and output the structural state identification result of the target special equipment through the acoustic-vibration collaborative identification model, and specifically includes: extracting acoustic spectrum features and vibration response features through the acoustic-vibration collaborative identification model based on the effective modal signal data; performing weighted fusion on the acoustic spectrum features and the vibration response features in a cross-channel attention fusion structure, and generating a joint feature vector; calculating a target acoustic-vibration risk score value through the acoustic-vibration collaborative identification model according to the joint feature vector; determining a size relationship between the target acoustic-vibration risk score value and a preset state threshold; if the target acoustic-vibration risk score value is greater than the preset state threshold, outputting the acoustic-vibration detection identification result as an abnormal acoustic-vibration detection identification result; and if the target acoustic-vibration risk score value is less than or equal to the preset state threshold, outputting the acoustic-vibration detection identification result as a normal acoustic-vibration detection identification result.

[0104] In a possible implementation, the defect type state label includes a cavity defect type state label and a delamination defect type state label, and the processing module 22 is configured to, after inputting the reconstructed effective modal signal data into the acoustic-vibration collaborative identification model and outputting the structural state identification result of the target special equipment through the acoustic-vibration collaborative identification model, further include: obtaining first training sample data corresponding to the delamination defect type state label, and obtaining second training sample data corresponding to the delamination defect type state label; constructing a defect type discrimination sub-model based on the first training sample data and the second training sample data, and integrating the defect type discrimination sub-model into the acoustic-vibration collaborative identification model to form a multi-task identification structure; inputting the reconstructed effective modal signal data into the defect type discrimination sub-model, and outputting a defect type probability distribution corresponding to the target special equipment based on the defect type discrimination sub-model; and outputting a target defect type corresponding to the target special equipment through the defect type discrimination sub-model based on the defect type probability distribution.

[0105] Optionally,

[0106] It should be noted that the apparatus provided in the above examples divides the internal structure of the device into different functional modules to complete all or part of the functions described above when realizing the functions of the apparatus. In actual applications, the functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above examples belong to the same concept, and the specific implementation process is detailed in the method embodiments, which will not be described here.

[0107] The present application also provides an electronic device. Referring to Figure 3 , Figure 3is a structural schematic diagram of an electronic device provided by an embodiment of the present application. The electronic device can include at least one processor 301, at least one communication bus 302, a user interface 303, at least one network interface 304, and a memory 305.

[0108] The communication bus 302 is configured to realize connection and communication between the components.

[0109] The user interface 303 can include a display and a camera. Optionally, the user interface 303 can further include a standard wired interface and a wireless interface.

[0110] The network interface 304 can optionally include a standard wired interface and a wireless interface (such as a WI-FI interface).

[0111] The processor 301 can include one or more processing cores. The processor 301 is connected to various parts of the server through various interfaces and lines, and performs various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 305, and calling data stored in the memory 305. Optionally, the processor 301 can be implemented in at least one of a hardware form of a digital signal processing (DSP), a field-programmable gate array (FPGA), and a programmable logic array (PLA). The processor 301 can be integrated with a combination of one or more of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. The CPU is mainly used to process an operating system, a user interface, and an application program. The GPU is used to render and draw the content to be displayed on the display. The modem is used to process wireless communication. It can be understood that the above-mentioned modem can also not be integrated into the processor 301, but can be implemented by a separate chip.

[0112] The memory 305 may include random access memory (RAM) or read-only memory. Optionally, the memory 305 may include a non-transitory computer-readable storage medium. The memory 305 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 305 may also be at least one storage device located remotely from the aforementioned processor 301. (Refer to...) Figure 3 The memory 305, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an anti-interference acoustic and vibration detection application.

[0113] exist Figure 3 In the illustrated electronic device, the user interface 303 is primarily used to provide an input interface for the user and acquire user input data; while the processor 301 can be used to call the anti-interference acoustic vibration detection application stored in the memory 305. When executed by one or more processors 301, the electronic device performs one or more of the methods described in the above embodiments. It should be noted that, for the foregoing method embodiments, for the sake of simplicity, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0114] This application also provides a computer-readable storage medium storing instructions. When executed by one or more processors, these instructions cause an electronic device to perform one or more of the methods described in the above embodiments.

[0115] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0116] In several embodiments provided in the present application, it should be understood that the disclosed apparatus can be implemented in other manners. For example, the division of the apparatus embodiments is merely illustrative, and the units can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0117] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0118] In addition, the functional units in each embodiment of the present application can be integrated into a processing unit, or each unit can be physically present separately, or two or more units can be integrated into one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0119] If the integrated unit is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the methods in the embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, magnetic disk or optical disk, and various program codes that can be stored.

[0120] The above is merely exemplary embodiments of the present application, and cannot limit the scope of the present application. Any equivalent changes and modifications made in accordance with the teachings of the present application are still within the scope of the present application.

[0121] The present application is intended to cover any variations, uses or adaptive changes of the present application, which follow the general principles of the present application and include common knowledge or conventional technical means in the technical field of the present application not disclosed in the present application.

Claims

1. An anti-interference acoustic vibration detection method, characterized in that, The method comprises: acquiring a target special equipment and a device type corresponding to the target special equipment; acquiring, according to the device type, a multi-channel high-sensitivity sensing array combination corresponding to the device type from a preset inspection method database, one device type corresponding to one multi-channel high-sensitivity sensing array combination; acquiring, based on the multi-channel high-sensitivity sensing array combination, acoustic vibration signal data corresponding to the target special equipment; applying an improved empirical mode decomposition algorithm to the acoustic vibration signal data for modal decomposition, eliminating pseudo-modal components and reconstructing effective modal signal data; inputting the reconstructed effective modal signal data into an acoustic vibration collaborative recognition model, and outputting a structure state recognition result of the target special equipment through the acoustic vibration collaborative recognition model.

2. The method of claim 1, wherein, Before the acquiring of the target special equipment and the device type corresponding to the target special equipment, the method further comprises: classifying the target special equipment according to the structure and function attribute categories of special equipment, classifying the special equipment into a plurality of device types, the device type corresponding to the target special equipment being any one of the plurality of device types, and the plurality of device types including pressure-bearing equipment, hoisting equipment, elevator equipment, high-altitude operation equipment, pressure special environment equipment, rail transit equipment, and aerospace equipment.

3. The method of claim 2, wherein, Before the acquiring, according to the device type, of the multi-channel high-sensitivity sensing array combination corresponding to the device type from the preset inspection method database, the preset inspection method database needs to be constructed, specifically comprising: constructing, based on different device types, a correspondence between the device types and the multi-channel high-sensitivity sensing array combination according to a preset correspondence mode, the preset correspondence mode including sensor type, number of acquisition channels, sensor spatial arrangement mode, inter-channel synchronous sampling accuracy, and signal acquisition frequency band range configuration mode; storing the correspondence in the preset inspection method database to construct the preset inspection method database.

4. The method of claim 3, wherein, The multi-channel high-sensitivity sensing array combination includes one or more combinations of a miniature acceleration sensor array, a laser vibration measurement sensor array, an acoustic emission sensor array, a piezoelectric sensor array, a microwave vibration probe array, a magnetostrictive sensor array, a wideband capacitive microphone array, and a directional ultrasonic sensor array. When the device type is the pressure-bearing equipment, the acquiring, based on the multi-channel high-sensitivity sensing array combination, of acoustic vibration signal data corresponding to the target special equipment specifically comprises: using the acoustic emission sensor array, the piezoelectric sensor array, the miniature acceleration sensor array, and the directional ultrasonic sensor array as the multi-channel high-sensitivity sensing array combination; acquiring acoustic emission transient pulse data based on the acoustic emission sensor array, harmonic response signal data based on the piezoelectric sensor array, acceleration time-domain response data based on the miniature acceleration sensor array, and ultrasonic echo feature data based on the directional ultrasonic sensor array; The acoustic emission transient pulse data, the harmonic response signal data, the acceleration time domain response data, and the ultrasonic echo feature data are taken as the acoustic vibration signal data.

5. The method of claim 1, wherein, The improved empirical mode decomposition algorithm is applied to the acoustic vibration signal data for modal decomposition, and false modal components are removed to reconstruct valid modal signal data, and the method specifically comprises the following steps: End point extension processing and white noise superposition processing are performed on the acoustic vibration signal data to enhance the separability of the modal components corresponding to the acoustic vibration signal data; Set pair empirical mode decomposition operation is performed on the enhanced acoustic vibration signal data to obtain an initial intrinsic mode function component sequence; Modal characteristic parameter values corresponding to the initial intrinsic mode function component sequence are calculated by the improved empirical mode decomposition algorithm, and the modal characteristic parameter values include energy proportion values, frequency stability values, and instantaneous frequency fluctuation values; Based on the modal characteristic parameter values, the corresponding false modal components in the acoustic vibration signal data are identified and removed according to a preset false modal identification rule; The remaining modal components after removal are accumulated and reconstructed to generate the valid modal signal data.

6. The method of claim 1, wherein, The reconstructed valid modal signal data is input into an acoustic vibration collaborative identification model, and a structure state identification result of the target special equipment is output by the acoustic vibration collaborative identification model, and the method specifically comprises the following steps: Historical acoustic vibration signal data corresponding to the identified special equipment is obtained; The improved empirical mode decomposition algorithm is applied to the historical acoustic vibration signal data for modal decomposition, and false modal components are removed to reconstruct historical valid modal signal data; Multi-scale feature parameters are obtained based on the historical valid modal signal data, and the multi-scale feature parameters include frequency distribution feature parameters, energy entropy feature parameters, envelope variation coefficient parameters, and transient response amplitude change rate parameters; Device state labels corresponding to the multi-scale feature parameters are obtained, and a training sample set is constructed based on the multi-scale feature parameters and the device state labels, and the device state labels include structure integrity level labels, operating condition labels, defect type labels, numerical classification labels, and time sequence evolution labels; A multi-layer neural network structure is constructed based on the training sample set, and the neural network structure includes feature extraction layers, cross-channel attention fusion layers, and state identification output layers corresponding to acoustic channels and vibration channels, respectively; The neural network is trained in a supervised learning manner, and the acoustic vibration collaborative identification model is constructed.

7. The method of claim 1, wherein, The reconstructed valid modal signal data is input into an acoustic vibration collaborative identification model, and a structure state identification result of the target special equipment is output by the acoustic vibration collaborative identification model, and the method specifically comprises the following steps: Based on the valid modal signal data, acoustic spectral features and vibration response features are extracted by the acoustic vibration collaborative identification model; The acoustic spectral features and the vibration response features are weighted and fused in a cross-channel attention fusion structure to generate a joint feature vector; A target acoustic vibration risk score value is calculated by the acoustic vibration collaborative identification model according to the joint feature vector; The size relationship between the target acoustic vibration risk score value and a preset state threshold value is judged; If the target acoustic vibration risk score value is greater than the preset state threshold value, the acoustic vibration detection recognition result is output as an abnormal acoustic vibration detection recognition result. If the target acoustic vibration risk score value is less than or equal to the preset state threshold value, the acoustic vibration detection recognition result is output as a normal acoustic vibration detection recognition result.

8. The method of claim 6, wherein, The defect type state label includes a cavity defect type state label and a delamination defect type state label, and after the reconstructed effective modal signal data is input into the acoustic vibration cooperative recognition model and the structure state recognition result of the target special equipment is output by the acoustic vibration cooperative recognition model, the method further includes: obtaining first training sample data corresponding to the delamination defect type state label and obtaining second training sample data corresponding to the delamination defect type state label; constructing a defect type discrimination sub-model based on the first training sample data and the second training sample data, and integrating the defect type discrimination sub-model into the acoustic vibration cooperative recognition model to form a multi-task recognition structure; inputting the reconstructed effective modal signal data into the defect type discrimination sub-model, and outputting a defect type probability distribution corresponding to the target special equipment based on the defect type discrimination sub-model; outputting a target defect type corresponding to the target special equipment by the defect type discrimination sub-model based on the defect type probability distribution.

9. An anti-interference acoustic vibration detection device, characterized in that, The device includes an acquisition module and a processing module, wherein, The acquisition module is configured to acquire a target special equipment and a device type corresponding to the target special equipment; acquire a multi-channel high-sensitivity sensor array combination corresponding to the device type in a preset inspection method database according to the device type, one device type corresponding to one multi-channel high-sensitivity sensor array combination; and acquire acoustic vibration signal data corresponding to the target special equipment based on the multi-channel high-sensitivity sensor array combination; The processing module is configured to apply an improved empirical mode decomposition algorithm to the acoustic vibration signal data for modal decomposition, eliminate pseudo-modal components, and reconstruct effective modal signal data; input the reconstructed effective modal signal data into an acoustic vibration cooperative recognition model, and output a structure state recognition result of the target special equipment by the acoustic vibration cooperative recognition model.

10. An electronic device, comprising: The electronic device includes a processor, a communication bus, a user interface, a network interface, and a memory, the memory is configured to store instructions, the user interface and the network interface are configured to communicate with other devices, and the processor is configured to execute the instructions stored in the memory to enable the electronic device to perform the method of any one of claims 1 to 8.

Images