A beam factory vibration noise control method based on big data

Abstract

The application discloses a beam factory vibration noise control method based on big data, and particularly relates to the field of intelligent control of industrial noise, and is used for solving the problem that vibration noise in a dynamic shielding environment of a beam factory is difficult to accurately control. By establishing a global three-dimensional coordinate system and dividing a three-dimensional space grid body, collecting precast beam boundary contour point cloud data and real-time coordinate data of a vibrating device, generating a dynamic space topology matrix, reducing and flattening the dynamic space topology matrix to form a one-dimensional space feature vector and searching a target acoustic mapping parameter set, calculating a residual sound pressure level value of an air transmission state three-dimensional space grid body in a three-dimensional virtual coordinate system and generating a dynamic sound field distribution prediction graph, and then based on an out-of-standard coordinate node, a propagation shielding analysis result and a device operation analysis result, joint determination is completed, a target vibrating device is further locked and a corresponding sound source control instruction is output, thereby realizing tracking control of beam factory vibration noise.

Description

Technical Field

[0001] This invention relates to the field of intelligent industrial noise control, and more specifically, to a method for controlling vibration noise in a beam fabrication plant based on big data. Background Technology

[0002] As a production base for large precast components, the beam fabrication plant continuously generates strong mechanical radiation noise during concrete pouring and vibration. Unlike noise from typical fixed equipment, the storage, hoisting, transportation, and relocation of precast beams within the plant constantly alter the physical distribution of the area, leading to continuous changes in sound wave propagation paths, reflection interfaces, and obstruction relationships. As production progresses, previously connected propagation channels may be divided by newly added beams, and previously open spaces may quickly transform into complex sound fields with multiple reflections and localized focusing.

[0003] Existing noise prediction and control schemes mostly use static drawings, fixed measuring points, or planar boundaries for modeling. These methods have some reference value in scenarios where the spatial boundaries of the plant area change little, but they are prone to the following problems in the dynamic shielding environment of a beam fabrication plant: First, they cannot reflect the reconstruction of the propagation boundary caused by beam stacking and relocation in a timely manner; second, it is difficult to establish a stable correspondence between specific locations exceeding the standard and specific vibrating equipment; third, control decisions are often based solely on the result of exceeding the standard, lacking joint analysis of propagation shielding factors and equipment operating status, which can easily lead to inaccurate control or delayed response.

[0004] Therefore, there is a need for a beam fabrication plant vibration noise control method that can update the propagation boundary in real time as the spatial structure of the beam fabrication plant changes, can quickly call up matching propagation parameters, and can trace back the noise exceeding the standard results to the target vibration equipment. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a big data-based method for controlling vibration noise in beam fabrication plants. By mapping the current spatial occupancy status of the beam and the equipment position status to the same three-dimensional spatial reference, the method utilizes a historical working condition database to quickly match the propagation parameters corresponding to the current scenario. Furthermore, it combines the location of exceeding the standard, the degree of propagation obstruction, and the abnormal frequency operation of the equipment for joint judgment, thereby outputting control commands corresponding to the target vibration equipment.

[0006] To achieve the above objectives, the present invention provides the following technical solution: Step S1: Establish a global three-dimensional coordinate system and divide the three-dimensional spatial mesh. Based on the point cloud data of the precast beam boundary contour and the real-time coordinate data of the vibrating equipment, determine the entity occlusion state and the air transmission state, and generate a dynamic spatial topology matrix. Step S2: Dimensionally reduce and flatten the dynamic spatial topology matrix to generate a one-dimensional spatial feature vector. Input the one-dimensional spatial feature vector into the acoustic feature database to match the historical environmental spatial feature vector, extract the spatial acoustic wave reflectivity and medium transmittance, and generate the target acoustic mapping parameter set. Step S3: Map the real-time coordinate data to virtual sound-emitting nodes in the three-dimensional virtual coordinate system, calculate the residual sound pressure level of the three-dimensional spatial grid in the air transmission state based on the dynamic spatial topology matrix and the target acoustic mapping parameter set, and generate a dynamic sound field distribution prediction map. Step S4: Based on the dynamic sound field distribution prediction map, the target vibrating device is locked, and the first and second analysis quantities are extracted based on the dynamic spatial topology matrix and the real-time drive current waveform data of the target vibrating device. The first and second analysis quantities are input into the pre-training judgment module for joint judgment, and the corresponding sound source control command of the target vibrating device is output.

[0007] Furthermore, a global three-dimensional coordinate system is established based on the permanent measurement benchmark of the beam fabrication plant. The unit reference side length of the three-dimensional spatial grid is set. Point cloud data of the precast beam boundary contour is collected and real-time coordinate data of the vibration equipment is received. The boundary contour point cloud data is mapped to the three-dimensional spatial grid. The three-dimensional spatial grid containing boundary contour point cloud data is assigned the value of solid occlusion state, and the three-dimensional spatial grid not containing boundary contour point cloud data is assigned the value of air transmission state. The state marker of each three-dimensional spatial grid is written into the dynamic spatial topology matrix.

[0008] Furthermore, before mapping the boundary contour point cloud data to the three-dimensional spatial mesh, the boundary contour point cloud data is first reconstructed into a closed entity envelope to form the closed entity boundary of the corresponding precast beam. Then, the spatial relationship between the mesh center coordinates of each three-dimensional spatial mesh and the closed entity boundary is used as a supplementary judgment criterion. The three-dimensional spatial mesh located inside the closed entity boundary is also assigned to the entity occlusion state, so that the dynamic spatial topology matrix simultaneously represents the volume distribution of the entity boundary and the entity interior.

[0009] Furthermore, the dynamic spatial topology matrix is ​​flattened and reduced in dimension according to a fixed traversal order, so that the entity occlusion state and the air transmission state are written into the one-dimensional spatial feature vector with a unique position number. The historical environmental spatial feature vectors in the acoustic feature database are stored with the same grid division rules, the same flattening order and the same state encoding rules as the one-dimensional spatial feature vectors, so that each position of the vector corresponds to the same spatial orientation when matching.

[0010] Furthermore, the one-dimensional spatial feature vector is matched with each historical environmental spatial feature vector in the acoustic feature database one by one using cosine similarity. The vectors are sorted from largest to smallest according to their cosine similarity values. The historical environmental spatial feature vector at the top of the sort is used as the record corresponding to the current spatial pattern. Spatial acoustic wave reflectivity and medium transmittance are extracted synchronously from the same record as the historical environmental spatial feature vector at the top of the sort and written into the target acoustic mapping parameter set.

[0011] Furthermore, in a three-dimensional virtual coordinate system that maintains the same coordinate semantics as the global three-dimensional coordinate system, real-time coordinate data is written into the virtual sound-generating node of the corresponding vibrating equipment, and the center coordinate of the air-transmitting three-dimensional spatial grid is used as the virtual receiving node. Virtual sound wave rays are established from each virtual sound-generating node to each virtual receiving node, and then continuous adjacent physical occlusion three-dimensional spatial grids are extracted along each virtual sound wave ray and merged to form a set of physical penetration segments.

[0012] Furthermore, for each set of physical penetration segments, the spatial acoustic wave reflectivity and medium transmittance in the target acoustic mapping parameter set are called to continuously calculate the interface reflection loss of the virtual acoustic ray at the incident and exit positions of the physical penetration segment and the transmission attenuation inside the physical penetration segment. Then, based on the initial nominal sound power and propagation distance, the single-source residual sound intensity of each virtual receiving node is obtained. All single-source residual sound intensities are accumulated and converted into residual sound pressure level values ​​to generate a dynamic sound field distribution prediction map.

[0013] Furthermore, the first analytical quantity includes the cumulative value of the line-of-sight obstruction section of the direct sound wave; Based on the dynamic sound field distribution prediction map, spatial coordinates with residual sound pressure levels higher than the safety threshold are extracted as the set of noise exceeding coordinate nodes. The target vibrating equipment is locked according to the single-source residual sound intensity contribution of each vibrating equipment to each noise exceeding coordinate. A direct sound wave ray is established between the virtual sound-generating node corresponding to the target vibrating equipment and the target exceeding node. Then, the three-dimensional spatial mesh of intersecting entities that are continuously adjacent along the direct sound wave ray is connected and merged. The overall projection section after merging is calculated to obtain the cumulative value of the direct sound wave line-of-sight occlusion section.

[0014] Furthermore, the second analytical quantity includes the cumulative value of the line-of-sight obstruction section of the direct sound wave; The real-time drive current waveform data of the target vibrating equipment is collected synchronously. The frequency component with the highest amplitude in the design vibration response frequency band is extracted from the real-time drive current waveform data within the stable operation window as the frequency corresponding to the main frequency characteristic peak. The difference between the frequency corresponding to the main frequency characteristic peak and the standard rated vibration frequency of the target vibrating equipment is calculated to obtain the cumulative value of the direct sound wave line-of-sight obstruction section.

[0015] Furthermore, the accumulated value of the direct line-of-sight obstruction section and the accumulated value of the direct line-of-sight obstruction section are written into the same two-dimensional input vector in a fixed order, and the two-dimensional input vector is input into the pre-training judgment module. The pre-training judgment module is a support vector regression model with radial basis kernel function deployed. The support vector regression model outputs the intervention urgency coefficient. Then, based on the correspondence between the intervention urgency coefficient and the stepped tolerance threshold group, the sound source control command signal that maintains the original operating frequency, the sound source control command signal that reduces the drive power supply frequency, or the sound source control command signal that cuts off the drive power supply is output.

[0016] The technical effects and advantages of the present invention, a method for controlling vibration noise in beam fabrication plants based on big data: First, a dynamic spatial topology matrix is ​​used to uniformly represent the constantly changing physical occlusion environment of the beam fabrication plant, which can synchronously update the propagation boundary as the position of the precast beam changes. Secondly, by using historical environment retrieval, the target acoustic mapping parameter set that matches the current spatial pattern can be quickly obtained, avoiding the need to re-perform the complete acoustic inversion with each refresh and improving the response efficiency in continuous operation scenarios. Third, by incorporating the dynamic sound field distribution prediction results, the degree of propagation obstruction, and the degree of equipment malfunction into the same judgment chain, the noise exceeding the standard result can be directly traced back to the specific equipment, and control commands with clear engineering implications can be output. Attached Figure Description

[0017] Figure 1 A schematic diagram of the overall process of a beam fabrication plant vibration noise control method based on big data, provided for an embodiment of the present invention; Figure 2 This is a schematic diagram of the dynamic spatial topology matrix generation process in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the process of generating the target acoustic mapping parameter set in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the process of generating a dynamic sound field distribution prediction map in an embodiment of the present invention; Figure 5 This is a schematic diagram of the target vibrating device locking and first analytical quantity extraction process in an embodiment of the present invention; Figure 6 This is a schematic diagram of the second analytical quantity extraction, joint determination, and sound source control command output process in an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figures 1-6 This invention provides a method for controlling vibration noise in beam fabrication plants based on big data, comprising: Step S1: Establish a global three-dimensional coordinate system and divide the three-dimensional spatial mesh. Based on the point cloud data of the precast beam boundary contour and the real-time coordinate data of the vibrating equipment, determine the entity occlusion state and the air transmission state, and generate a dynamic spatial topology matrix. Step S2: Dimensionally reduce and flatten the dynamic spatial topology matrix to generate a one-dimensional spatial feature vector. Input the one-dimensional spatial feature vector into the acoustic feature database to match the historical environmental spatial feature vector, extract the spatial acoustic wave reflectivity and medium transmittance, and generate the target acoustic mapping parameter set. Step S3: Map the real-time coordinate data to virtual sound-emitting nodes in the three-dimensional virtual coordinate system, calculate the residual sound pressure level of the three-dimensional spatial grid in the air transmission state based on the dynamic spatial topology matrix and the target acoustic mapping parameter set, and generate a dynamic sound field distribution prediction map. Step S4: Based on the dynamic sound field distribution prediction map, the target vibrating device is locked, and the first and second analysis quantities are extracted based on the dynamic spatial topology matrix and the real-time drive current waveform data of the target vibrating device. The first and second analysis quantities are input into the pre-training judgment module for joint judgment, and the corresponding sound source control command of the target vibrating device is output. The first analysis quantity includes the cumulative value of the direct sound wave line-of-sight obstruction section, and the second analysis quantity includes the cumulative value of the direct sound wave line-of-sight obstruction section.

[0020] This invention takes the dynamic occlusion scenario of a beam fabrication plant as the processing object. First, it constructs a dynamic spatial topology matrix using the point cloud data of the precast beam boundary contour and the real-time coordinate data of the vibrating equipment. Then, it converts the dynamic spatial topology matrix into a one-dimensional spatial feature vector and extracts the target acoustic mapping parameter set from the acoustic feature database. Subsequently, it completes the mapping of virtual sound-generating nodes, the calculation of propagation paths, and the numerical determination of residual sound pressure levels in a three-dimensional virtual coordinate system. Finally, it outputs sound source control commands based on the dynamic sound field distribution prediction map, the propagation occlusion analysis results, and the equipment operation analysis results, forming a continuous input, processing, and output chain.

[0021] During the continuous pouring, moving, and storage of precast beams at the beam fabrication plant, the position, quantity, and stacking pattern of the precast beams are constantly changing. The location of the vibrating equipment also changes continuously with the construction rhythm. The original open space of the plant area is rapidly cut by a large number of newly added entities, and the sound wave propagation boundary is therefore in a state of constant reconstruction. At this time, if the spatial scene is still at the level of static drawings or local coordinates, there is no unified reference for the subsequent identification of propagation paths, reflection interfaces, and occlusion relationships. Therefore, step S1 needs to first convert the current entity distribution and equipment position into a calculable, inheritable, and referential spatial representation result.

[0022] S101: Establish a unified spatial benchmark for the plant area and complete the three-dimensional spatial mesh division. Establish a global three-dimensional coordinate system using the permanent measurement control points of the plant area as a unified reference. Preferably, the direction along the precast beam conveying direction can be defined as the longitudinal coordinate direction, the lateral unfolding direction of the platform can be defined as the lateral coordinate direction, and the elevation direction can be defined as the vertical coordinate direction. Then, determine the effective longitudinal length, effective lateral width, and effective vertical height based on the actual operating range of the plant area, and set the unit benchmark side length. The unit benchmark side length is used to limit the boundary scale of a single three-dimensional spatial mesh in three directions. Its selection principle is: not less than the scanning point distance so that a single mesh can carry continuous surface sampling information; not greater than the minimum structural thickness of the precast beam to be identified so that the web, top plate, and flange boundaries of the beam can be distinguished at the mesh level. In engineering, the unit benchmark side length can be selected from 0.2 meters to 1.0 meters based on the scanning point distance and minimum structural thickness, preferably 0.5 meters. After setting the side length, the effective space of the factory area is discretized into regular three-dimensional spatial grids, and each grid is assigned a unique vertical, horizontal and vertical number. At the same time, the grid center coordinates of each grid are recorded for subsequent entity occupation determination.

[0023] S102: Acquire boundary contour point clouds and simultaneously collect real-time coordinate data. After mesh generation, three-dimensional laser scanning devices are deployed on both sides of the platform, above the hoisting channel, and at the edge of the beam storage area to perform multi-view scanning on the outer surface of the precast beam. The original contour points obtained by each scanning device are located in the local coordinate domain. Rigid body registration is first performed according to the installation posture and position of the scanning device, and then they are uniformly mapped to the global three-dimensional coordinate system to form a boundary contour point cloud set. At the same time, the built-in positioning module of each vibrating device outputs the real-time coordinates of the vibrating device, forming a real-time coordinate dataset. To ensure that the boundary contour point cloud set and the real-time coordinate dataset describe the same physical moment, a synchronization allowable time difference is introduced. The synchronization allowable time difference is used to limit the maximum allowable deviation between the scanning sampling time and the positioning output time. Its value can be determined jointly based on the scanning refresh cycle, the positioning refresh cycle, and the control refresh cycle, preferably not greater than half of the shortest refresh cycle among the three. For example, when the laser scanning device refreshes every 1 second, the positioning module outputs every 0.2 seconds, and the control system performs a topology refresh every 1 second, the allowable time difference for synchronization can be set to 0.5 seconds. Only when the time difference between the point cloud acquisition time and the device coordinate output time does not exceed 0.5 seconds can the two enter the same round of topology refresh.

[0024] S103: Reconstruct the boundary contour point cloud into a closed solid envelope. The boundary contour point cloud set only reflects the discrete sampling positions of the outer surface of the precast beam. If the values ​​are directly assigned based on whether the contour points fall into the mesh, only the mesh near the surface can be identified, making it difficult to write the internal solid volume of the beam into the dynamic spatial topology matrix. Therefore, before performing occupancy determination, the boundary contour point cloud set is reconstructed into a closed solid envelope. This reconstruction can be achieved using existing 3D point cloud closed reconstruction algorithms, such as Poisson reconstruction, α-shape reconstruction, or other methods that can obtain closed triangular surfaces. After reconstruction, a closed solid boundary is formed, which is used to completely represent the outer envelope solid boundary of all precast beams. Through this processing, any subsequent 3D spatial mesh can be determined as either a solid occlusion state or an air transmission state based on its spatial relationship with the closed solid boundary.

[0025] S104: Determine whether a 3D spatial mesh is in a solid occlusion state. After the closed solid boundary is formed, perform boundary occupancy and interior occupancy checks on each 3D spatial mesh. For boundary occupancy checks, the shortest distance from the mesh center coordinates to the closed solid boundary is used as the criterion. If the shortest distance is not greater than the radius of the circumscribed sphere of the mesh, the mesh is recorded as solid occupancy. For interior occupancy checks, a test ray is emitted from the mesh center coordinates along a preset direction, and the number of intersections between the test ray and the closed solid boundary is counted. If the number of intersections is odd, it indicates that the mesh center is located inside the closed solid boundary, and the mesh is also recorded as solid occupancy. If neither boundary occupancy nor interior occupancy checks are satisfied, the mesh is recorded as air-transmitting. This allows for the simultaneous identification of the area near the beam surface and the internal volume of the beam.

[0026] S105: Form and output the dynamic spatial topology matrix. After all 3D spatial meshes have completed state assignment, the state markers corresponding to each mesh are written into the 3D matrix elements according to a fixed arrangement of vertical, horizontal, and vertical indices. Entity occlusion states are uniformly recorded as entity occupancy markers, and air transmission states are uniformly recorded as air transmission markers, thus forming the dynamic spatial topology matrix. It should be noted that the dynamic spatial topology matrix is ​​used to characterize the spatial distribution of all entity boundaries, internal volumes, and air propagation areas within the beam fabrication plant at the refresh time. The real-time coordinate dataset is used to characterize the positional distribution of all vibrating equipment at the same refresh time. Both are parallel outputs of step S1. The dynamic spatial topology matrix is ​​used in subsequent spatial feature retrieval and propagation calculation processes, while the real-time coordinate dataset is used in subsequent virtual sound node mapping processes.

[0027] After step S1 is completed, the current moment's precast beam boundary, internal occupancy relationship, and vibratory equipment position have been uniformly written into the dynamic spatial topology matrix and real-time coordinate dataset. The dynamic spatial topology matrix represents the distribution of the entity occlusion state and air transmission state of all three-dimensional spatial meshes, and the real-time coordinate dataset represents the spatial position of each vibratory equipment at the same moment, providing a consistent coordinate basis and clear data source for subsequent spatial feature retrieval and virtual sound source mapping.

[0028] Step S1 has fixed the spatial occlusion pattern of the beam fabrication plant at the current moment as a dynamic spatial topology matrix. However, this result still belongs to a three-dimensional discrete structure, making it difficult to directly compare it with historical conditions quickly. At the same time, the propagation boundary parameters required in step S3 are not directly given by the dynamic spatial topology matrix, but need to be retrieved from the historical environment to obtain the propagation attributes that are closest to the current occlusion pattern. Based on this relationship, step S2 needs to first transcribe the dynamic spatial topology matrix into a spatial code with a unified order, and then establish a matching relationship with the acoustic feature database, so that the current scene can enter the searchable historical environment comparison process.

[0029] S201: Sequence Expansion and Query Key Generation of the Dynamic Spatial Topology Matrix. Since the dynamic spatial topology matrix is ​​a three-dimensional discrete structure, it is not suitable for direct, position-by-position comparison with historical records. Therefore, the current number of vertical, horizontal, and triangular grid cells is read first, and all grid cells are expanded into a one-dimensional spatial feature vector according to a fixed traversal order. Preferably, a rule of prioritizing vertical sequence numbers, followed by horizontal sequence numbers, and finally vertical sequence numbers can be used to write entity occupancy markers or air penetration markers into a single-column sequence cell by cell. This one-dimensional spatial feature vector is the query key for the current scene, and each position in the vector uniquely corresponds to a three-dimensional spatial grid cell position.

[0030] S202: Record structure, database construction method, and consistency constraints of the acoustic feature database. To ensure spatial comparability of historical retrieval, each historical record in the acoustic feature database is stored using the same grid partitioning rules, the same flattening order, and the same state coding rules as in step S1. Each historical record includes at least three parts: historical environmental spatial feature vector, spatial acoustic reflectivity, and medium transmittance. The historical environmental spatial feature vector records the spatial occupancy distribution under a specific historical condition, and its generation method is consistent with the current query primary key; that is, it first performs grid occupancy judgment on the point cloud and entity envelope at the historical moment, and then flattens them in a fixed order. The spatial acoustic reflectivity characterizes the equivalent reflection degree of the entity boundary corresponding to the historical condition; the medium transmittance characterizes the equivalent transmission degree of the entity penetration corresponding to the historical condition. All three are bound and stored with the same historical record number.

[0031] Furthermore, spatial acoustic reflectivity and media transmittance can be obtained through acoustic calibration under historical operating conditions. Specifically, after forming a historical environmental spatial feature vector for a certain historical operating condition, several calibration measurement points can be preset within the plant area, and calibration sound sources can be deployed at one or more typical sound source locations, or a vibrating device in stable operation can be selected as the calibration sound source. Subsequently, actual sound level data at each calibration measurement point are collected, and a propagation path is constructed using the current historical environmental spatial feature vector. Then, by gradually adjusting the spatial acoustic reflectivity and media transmittance, the difference between the predicted sound level calculated from the historical environmental spatial feature vector and these two parameters and the measured sound level at the calibration measurement point falls within a preset error range, preferably within ±3 dB. When the error requirement is met, the set of spatial acoustic reflectivity and media transmittance is written into the database as the calibration result of the corresponding historical record. Thus, it can be seen that the spatial acoustic reflectivity and media transmittance in this application are not abstract parameters set out of thin air, but equivalent propagation parameters formed after being constrained by measured data and corresponding one-to-one with a specific historical spatial pattern.

[0032] S203: Similarity retrieval between the query primary key and historical environmental spatial feature vectors. After inputting the current query primary key into the acoustic feature database, the historical environmental spatial feature vectors in the historical records are read one by one, and the query primary key and each historical environmental spatial feature vector are compared using cosine similarity calculation. The higher the cosine similarity, the more consistent the spatial occupancy direction at the current refresh time is with the corresponding historical conditions. After comparing all historical records, they are sorted from high to low according to cosine similarity, with the record at the top of the sorted list being given priority as a candidate matching record. To avoid forcibly calling unsuitable historical parameters when the similarity is insufficient, an effective retrieval threshold can be set. The effective retrieval threshold can be determined based on the historical validation set. Specifically, it can be done as follows: on several known historical samples, different lower limits of similarity are used as screening thresholds, and the error between the predicted sound level and the measured sound level corresponding to the called record is compared. The lowest similarity that allows most validation samples to meet the preset error requirement is selected as the effective retrieval threshold. For example, when the similarity is not lower than 0.85, more than 80% of the validation samples can control the prediction error within ±3 dB, so 0.85 can be set as the effective retrieval threshold.

[0033] When the first record in the sorted list does not meet the valid retrieval threshold, or when the current query primary key consists entirely of air transmission markers, the retrieval process does not directly use the first record in the sorted list, but instead switches to the baseline open field record number. The baseline open field record number corresponds to the unobstructed factory area condition. It can be generated by performing an independent scan within the factory area under unobstructed conditions or with only fixed infrastructure, forming an open field historical environmental spatial feature vector. After calibrating the spatial acoustic reflectivity and medium transmittance using the same acoustic calibration method as described above, it is stored in the database. The baseline open field record number is retained as a fallback record in the long term to ensure that the system can still obtain a usable set of target acoustic mapping parameters in extremely open conditions or when not finding a sufficiently similar historical record.

[0034] S204: Synchronous Extraction and Output of the Target Acoustic Mapping Parameter Set. Once the valid matching record number is determined, the historical environmental spatial feature vector itself is no longer processed. Instead, the corresponding historical record unit is entered based on the record number, and the spatial acoustic wave reflectivity and medium transmittance are read synchronously and written into the same parameter container to form the target acoustic mapping parameter set. This process maintains the same source extraction and avoids the situation of extracting parameters from different historical records and then splicing them together.

[0035] After step S2 is completed, the dynamic spatial topology matrix has been reduced in dimension and flattened into a one-dimensional spatial feature vector. This vector is then matched and sorted against historical environmental spatial feature vectors in the acoustic feature database. The spatial acoustic reflectivity and medium transmittance bound to the first record in the sorted list are simultaneously extracted and written into the target acoustic mapping parameter set. At this point, the current beam fabrication plant's obstruction pattern has obtained a corresponding source of equivalent propagation parameters, providing clear propagation attribute inputs for subsequent sound field calculations.

[0036] Step S2 has provided the target acoustic mapping parameter set, and step S1 has provided the dynamic spatial topology matrix and real-time coordinate dataset. At this point, the propagation boundary, sound source location, and equivalent propagation parameters are complete. However, these results remain at the parameter and structural levels and have not yet been converted into noise values ​​corresponding to spatial locations. What the beam fabrication plant truly needs to identify is which air propagation areas are experiencing what level of residual sound pressure. Therefore, step S3 requires writing the existing inputs into a three-dimensional virtual coordinate system and performing diffusion and penetration attenuation calculations grid by grid along the propagation path, so that the spatial sound field distribution forms a quantitative result that can be traversed.

[0037] S301: Establishment of Virtual Sound Generating Nodes and Setting of Lower Boundary for Effective Propagation Distance. To ensure the spatial semantics in step S1 are carried over to the sound field prediction process, the three-dimensional virtual coordinate system adopts the same origin and coordinate direction as the global three-dimensional coordinate system. The positions of each vibrating device in the real-time coordinate dataset are directly written into the three-dimensional virtual coordinate system, forming corresponding virtual sound generating nodes. Simultaneously, the initial nominal sound power of each vibrating device under rated vibration operation is read from the acoustic calibration file, factory test record, or on-site rated operating condition calibration results of the vibrating device, and bound to the corresponding virtual sound generating node. Considering that the vibrating device may be adjacent to some air-transmitting mesh, directly using a minimum propagation distance for diffusion conversion could easily lead to non-physical amplification; therefore, a minimum propagation radius is set. The minimum propagation radius represents the lower bound of the shortest distance allowed for sound source propagation calculation, preferably not less than one unit reference side length, but can also be determined according to the equivalent radius of the device shell. For any subsequent virtual sound wave ray, the larger of the actual propagation distance and the minimum propagation radius is used as the effective propagation distance during diffusion conversion.

[0038] S302: Virtual Receiving Node Establishment and Solid Penetration Segment Set Extraction. The center coordinates of all air-transmitting 3D spatial meshes are defined as virtual receiving nodes, which are the calculation points for residual sound pressure levels. Then, a virtual acoustic ray is established from each virtual sound-emitting node to each virtual receiving node, and the length of this line is recorded as the basic propagation distance. The continuous meshes marked as solid occlusion states in the dynamic spatial topology matrix are scanned along the virtual acoustic ray. Adjacent continuous solid occlusion state meshes without air-transmitting mesh intervals are merged into a solid penetration segment. All solid penetration segments together constitute the solid penetration segment set of this ray. The purpose of this process is to eliminate the artificial boundaries formed within the same precast beam due to mesh discretization, ensuring that actual reflection only occurs at the points of entry and exit from the solid, without repeated calculations at each mesh boundary within the beam.

[0039] S303: Target Acoustic Mapping Parameter Set Invocation and Single-Source Residual Sound Intensity Calculation. The target acoustic mapping parameter set consists of spatial acoustic reflectivity and medium transmittance. Spatial acoustic reflectivity represents the equivalent energy proportion of sound waves that are reflected and no longer propagate along the original direction when they reach the entity boundary. Medium transmittance represents the proportion of forward energy retained by the sound wave after passing through one unit reference side length inside the entity. For any entity penetration segment, an interface reflection loss is first applied at the incident position. Then, the medium transmittance is repeatedly applied segment by segment according to the entity propagation length, divided into several unit reference side length intervals. Finally, an interface reflection loss is applied at the exit position. If the same virtual acoustic ray passes through multiple entity penetration segments, the forward energy retention results of each entity penetration segment are continuously connected according to the propagation sequence to obtain the total forward energy retention factor of the virtual acoustic ray propagating from the virtual sound-emitting node to the virtual receiving node.

[0040] For example, if the spatial acoustic wave reflectivity corresponding to a certain historical record is 30%, it means that approximately 30% of the forward energy is reflected each time it passes through a solid boundary; and the medium transmittance is 80%, meaning that the acoustic wave retains 80% of its forward energy after passing through a unit reference side length inside the solid. If a ray passes through a solid penetration segment with a length of 3 reference sides, it retains 70% of its forward energy at the incident interface, 80% of its forward energy is retained three times consecutively inside the solid, and another 70% of its forward energy is retained at the exit interface, ultimately leaving approximately one-quarter of the original forward energy. If the ray subsequently passes through a second solid penetration segment, the same rules are applied sequentially based on the results of the previous segment.

[0041] After calculating the total forward energy retention factor, the initial nominal sound power is then distributed to the propagation area corresponding to the effective propagation distance according to the spherical diffusion law to obtain the propagation sound intensity under unobstructed diffusion conditions. This intensity is then multiplied by the total forward energy retention factor to obtain the single-source residual sound intensity from any virtual sound-emitting node to any virtual receiving node.

[0042] S304: Multi-source residual sound pressure level conversion and dynamic sound field distribution prediction map generation. Since multiple vibrating devices in a beam fabrication plant are typically driven independently by different workstations, maintaining a stable, fixed phase relationship over a long period is difficult. Therefore, the multi-source contributions received by the same virtual receiving node can be processed by directly accumulating the sound intensity. The single-source residual sound intensities of all virtual sound-emitting nodes pointing to the same virtual receiving node are added one by one to obtain the total residual sound intensity. Then, air characteristic impedance is introduced to convert the total residual sound intensity into the corresponding sound pressure level, obtaining the residual sound pressure level value at the virtual receiving node. Air characteristic impedance can be obtained in two ways: first, by reading data from on-site temperature and humidity sensors and air pressure sensors, combined with a built-in air parameter conversion table to determine air density and sound velocity, and then obtaining the air characteristic impedance; second, by using preset operating condition reference values ​​when no environmental sensors are available. For example, under normal temperature and pressure, the air characteristic impedance corresponding to an air density of approximately 1.2 kg / m³ and a sound velocity of approximately 340 m / s can be used as the default value, and this parameter is updated when the temperature change exceeds the preset range. After completing the numerical calculation of the residual sound pressure level of all air-transmitting grid cells, the spatial coordinates of each virtual receiving node and the corresponding residual sound pressure level value are bound and written into the coordinate sound level recording unit. All coordinate sound level recording units together constitute a dynamic sound field distribution prediction map.

[0043] After step S3 is completed, the real-time coordinate dataset has been mapped to virtual sound-emitting nodes, the dynamic spatial topology matrix has been transformed into propagation boundaries, the target acoustic mapping parameter set has been used to calculate the energy attenuation on each propagation path, and the residual sound pressure level corresponding to the three-dimensional spatial grid of the air transmission state has been obtained one by one and bound to the spatial coordinates to form a dynamic sound field distribution prediction map. At this point, the noise distribution in the air propagation area of ​​the beam fabrication plant is expressed by discrete coordinates and corresponding sound levels.

[0044] Step S3 has generated a dynamic sound field distribution prediction map, allowing direct identification of locations exceeding the standard in the air propagation area of ​​the beam fabrication plant. However, the result of exceeding the standard alone cannot distinguish whether the noise problem mainly originates from propagation anomalies caused by complex obstructions or from the abnormal frequency operation of the target vibrating equipment itself. If the control action lacks this distinction, it is easy to make unbalanced handling between propagation-side anomalies and equipment-side anomalies. Therefore, step S4 needs to continue to trace back to the target vibrating equipment based on the locations exceeding the standard, and extract the spatial propagation-side analysis quantity and the equipment operation-side analysis quantity respectively before entering the unified judgment process.

[0045] S401: Noise Exceeding Coordinate Node Extraction and Target Vibration Equipment Locking. After the dynamic sound field distribution prediction map is formed, safety thresholds are first set based on the operating period, environmental protection requirements, and plant management standards. Preferably, a threshold table can be pre-established, distinguishing different thresholds at least according to the operating period and target area category. The area to which the current node belongs and the current time period jointly determine the safety threshold used in this round. For example, an 85 dB control value can be used for the daytime area where personnel stay in the plant, while the corresponding control value can be calculated according to the local emission limit for nodes in the sensitive direction near the plant boundary. Subsequently, the dynamic sound field distribution prediction map is traversed, and all spatial coordinates with residual sound pressure levels higher than the current safety threshold are extracted as the set of noise exceeding coordinate nodes.

[0046] For each spatial coordinate in the set of noise-exceeding coordinate nodes, retrieve the single-source residual sound intensity results already saved in step S3, compare the single-source residual sound intensity contributions of all vibrating devices to that spatial coordinate, and lock the vibrating device with the highest contribution as the target vibrating device for that node. For example, if the single-source residual sound intensity contributions of devices A, B, and C at a certain noise-exceeding node are A highest, B second highest, and C lowest, then device A is locked as the target vibrating device for that node. After completing point-by-point locking, the noise-exceeding nodes belonging to the same target vibrating device are aggregated into a subset of device-associated noise-exceeding nodes, and the spatial coordinate with the highest residual sound pressure level value is selected from this subset as the target noise-exceeding node, thereby establishing a unique subsequent analysis entry point for each target vibrating device.

[0047] S402: Establishment of Direct Sound Wave Ray and Calculation of Accumulated Value of Direct Sound Wave Line-of-Sight Obstruction Section. After determining the target vibrating equipment and the target exceeding the standard node, the virtual sound-emitting node corresponding to the target vibrating equipment and the center coordinates of the target exceeding the standard node are directly connected in the three-dimensional virtual coordinate system to form a direct sound wave ray. The dynamic space topology matrix is ​​traversed along this direct sound wave ray, and all the physical obstruction state meshes that intersect with the ray's origin space are extracted to form an intersecting physical mesh set. Since the same precast beam is often discretized into multiple continuous meshes in the dynamic space topology matrix, if the projected area is accumulated cell by cell, the artificial mesh boundary will be repeatedly counted. Therefore, the intersecting physical mesh set is first connected and merged, and the intersecting physical meshes that are adjacent to each other and have no air transmission state interval in between are merged into a line-of-sight obstruction segment mesh group. Next, for each line-of-sight obstruction segment mesh group, the overall outer envelope is extracted, and the maximum outer contour projection area of ​​this overall outer envelope is calculated on the normal plane perpendicular to the direct sound wave ray. Finally, the maximum outer contour projection areas of all line-of-sight obstruction segment mesh groups are summed item by item to obtain the cumulative value of the direct sound wave line-of-sight obstruction section. The larger this value, the more complex the overall obstruction entity between the target device and the target exceeding the standard node is, and the more likely the current exceeding the standard result is to be affected by complex obstruction, indirect propagation, and local focusing.

[0048] S403: Real-time drive current waveform spectrum analysis and source equipment frequency deviation amplitude extraction. After determining the spatial propagation analysis quantity, the operating status information is further extracted from the target vibrating equipment. First, the real-time drive current waveform data corresponding to the current refresh time of the target vibrating equipment is collected, and then a stable operating time window is extracted from it. The stable operating time window can be defined as a time period in which the root mean square fluctuation of the current does not exceed a preset proportion within a continuous preset duration, for example, the root mean square fluctuation of the current does not exceed 10% within 2 consecutive seconds. Then, DC component removal processing is performed on the stable operating time window to eliminate the influence of static bias on the identification of spectrum peaks.

[0049] After preprocessing, spectral analysis is performed on the real-time drive current waveform data within the stable operation window, preferably using Fast Fourier Transform to obtain the spectral amplitude distribution. To avoid the power supply fundamental, low-frequency drift, or higher-order harmonics masking the dominant frequency truly reflecting the vibration state, the search range is limited to the design vibration response frequency band of the target vibrating equipment. The design vibration response frequency band and the standard rated vibration frequency can be determined from the equipment manual, nameplate, factory test records, or on-site calibration results, and are stored in the equipment parameter file. The frequency component with the highest amplitude is extracted from the design vibration response frequency band, and the frequency value corresponding to this component is defined as the frequency corresponding to the peak characteristic of the dominant frequency. The difference between the frequency corresponding to the peak characteristic of the dominant frequency and the standard rated vibration frequency is then calculated, and the absolute value of the difference is taken to obtain the deviation amplitude of the source equipment's abnormal frequency operation. The larger this value, the more obvious the deviation of the current operating frequency from the rated vibration state, and the higher the possibility of the equipment being overloaded, abnormally coupled, or instable in vibration state.

[0050] S404: Two-dimensional input vector writing, support vector regression determination, and sound source control command output. Once the accumulated value of the direct sound wave line-of-sight obstruction section and the deviation amplitude of the source equipment's inter-frequency operation have been obtained, standard processing is first performed on these two quantities. Then, they are written into a two-dimensional input vector in a fixed order. The first dimension contains the accumulated value of the direct sound wave line-of-sight obstruction section, and the second dimension contains the deviation amplitude of the source equipment's inter-frequency operation. Standard processing can employ interval normalization, that is, compressing the two quantities in the training samples to a uniform numerical range. The on-site input uses the same set of upper and lower bound parameters determined during the training phase to complete the conversion.

[0051] The pre-training decision module preferably employs a support vector regression model with a radial basis function kernel. For example, 3000 historical samples can be collected. Each historical sample includes the cumulative value of the line-of-sight obstruction section of the direct sound wave, the deviation amplitude of the source equipment operating at different frequencies, and the corresponding level of urgency for manual intervention. The level of urgency for manual intervention is then converted into a continuous regression target value. The sample set can be divided into 2100 training samples, 450 validation samples, and 450 test samples. Before training, abnormal missing samples are removed, duplicate samples are deduplicated, and both input values ​​are normalized to the same scale. Model hyperparameters can be determined through grid search combined with cross-validation.

[0052] The intervention urgency coefficient output by the support vector regression model is used to comprehensively characterize the urgency of handling the noise exceedance result of the current target vibrating equipment. Preferably, the intervention urgency coefficient can be normalized to the range of 0 to 1, with a larger value indicating a greater need for immediate and stronger control actions. Then, intermediate and advanced switching thresholds are preset, where the intermediate switching threshold represents the boundary between "maintaining operation" and "frequency reduction control," and the advanced switching threshold represents the boundary between "frequency reduction control" and "shutdown control." The two thresholds can be determined jointly through historical sample playback and on-site trial operation, that is, by comparing the excessive shutdown rate, missed control rate, and noise reduction effect under different threshold combinations, and selecting the threshold group with the best overall effect. For example, the intermediate switching threshold can be set to 0.40, and the advanced switching threshold can be set to 0.75.

[0053] The final control rules are as follows: When the intervention urgency coefficient is lower than the intermediate switching threshold, a sound source control command signal is output to maintain the original operating frequency. This signal indicates that the current drive frequency will remain unchanged and monitoring will continue. When the intervention urgency coefficient is not lower than the intermediate switching threshold and is lower than the advanced switching threshold, a sound source control command signal is output to reduce the drive power supply frequency. This signal indicates that the inverter output frequency or drive power supply frequency will be reduced by a preset percentage, preferably 5% to 15%, and resampled and evaluated after a preset time. When the intervention urgency coefficient is not lower than the advanced switching threshold, a sound source control command signal to cut off the drive power supply is output. This signal indicates that the machine will stop immediately and an alarm will be output. Operation will resume after manual review.

[0054] Through the above steps, the dynamic sound field distribution prediction map can be further transformed into a set of coordinate nodes exceeding the noise standard, the target vibration equipment, the propagation side analysis quantity, the equipment operation side analysis quantity, and the corresponding control commands. Under the condition that the spatial obstruction relationship in the beam fabrication plant is constantly changing, the vibration noise can be dynamically tracked and controlled in a targeted manner.

[0055] After step S4 is completed, the dynamic sound field distribution prediction map has been further transformed into a set of noise exceeding coordinate nodes, target vibrating equipment, spatial propagation side analysis quantity, equipment operation side analysis quantity, and corresponding sound source control commands. The exceeding location, propagation obstruction relationship, and equipment frequency status are included in the same judgment chain. The final output processing result directly corresponds to the operation control of the target vibrating equipment, providing a clear execution object and judgment basis for the vibration noise control of the beam fabrication plant at the current moment.

[0056] Specifically, the above are merely preferred embodiments of this application and are not intended to limit this application.

[0057] In the description of this specification, references to terms such as "an embodiment," "example," and "specific example" indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0058] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A method for controlling vibration noise in a beam fabrication plant based on big data, characterized in that, Including the following steps: Step S1: Establish a global three-dimensional coordinate system and divide the three-dimensional spatial mesh. Based on the point cloud data of the precast beam boundary contour and the real-time coordinate data of the vibrating equipment, determine the entity occlusion state and the air transmission state, and generate a dynamic spatial topology matrix. Step S2: Dimensionally reduce and flatten the dynamic spatial topology matrix to generate a one-dimensional spatial feature vector. Input the one-dimensional spatial feature vector into the acoustic feature database to match the historical environmental spatial feature vector, extract the spatial acoustic wave reflectivity and medium transmittance, and generate the target acoustic mapping parameter set. Step S3: Map the real-time coordinate data to virtual sound-emitting nodes in the three-dimensional virtual coordinate system, calculate the residual sound pressure level of the three-dimensional spatial grid in the air transmission state based on the dynamic spatial topology matrix and the target acoustic mapping parameter set, and generate a dynamic sound field distribution prediction map. Step S4: Based on the dynamic sound field distribution prediction map, the target vibrating device is locked, and the first and second analysis quantities are extracted based on the dynamic spatial topology matrix and the real-time drive current waveform data of the target vibrating device. The first and second analysis quantities are input into the pre-training judgment module for joint judgment, and the corresponding sound source control command of the target vibrating device is output.

2. The method for controlling vibration noise in a beam fabrication plant based on big data as described in claim 1, characterized in that, Step S1 includes: A global three-dimensional coordinate system is established based on the permanent measurement benchmark of the beam fabrication plant. The unit reference side length of the three-dimensional spatial grid is set. Point cloud data of the precast beam boundary contour is collected and real-time coordinate data of the vibration equipment is received. The boundary contour point cloud data is mapped to the three-dimensional spatial grid. The three-dimensional spatial grid containing the boundary contour point cloud data is assigned the value of solid occlusion state, and the three-dimensional spatial grid not containing the boundary contour point cloud data is assigned the value of air transmission state. The state flag of each three-dimensional spatial grid is written into the dynamic spatial topology matrix.

3. The method for controlling vibration noise in a beam fabrication plant based on big data, as described in claim 2, is characterized in that... Step S1 also includes: Before mapping the boundary contour point cloud data to the three-dimensional spatial mesh, the boundary contour point cloud data is first reconstructed into a closed entity envelope to form the closed entity boundary of the corresponding precast beam. Then, the spatial relationship between the mesh center coordinates of each three-dimensional spatial mesh and the closed entity boundary is used as a supplementary judgment criterion. The three-dimensional spatial mesh located inside the closed entity boundary is also assigned to the entity occlusion state, so that the dynamic spatial topology matrix can simultaneously represent the volume distribution of the entity boundary and the entity interior.

4. The method for controlling vibration noise in a beam fabrication plant based on big data as described in claim 1, characterized in that, Step S2 includes: The dynamic spatial topology matrix is ​​flattened and dimension-reduced according to a fixed traversal order. The entity occlusion state and the air transmission state are written into the one-dimensional spatial feature vector with unique position numbers. The historical environmental spatial feature vectors in the acoustic feature database are stored with the same grid division rules, the same flattening order and the same state encoding rules as the one-dimensional spatial feature vectors, so that each position of the vector corresponds to the same spatial orientation when matching.

5. The method for controlling vibration noise in a beam fabrication plant based on big data, as described in claim 4, is characterized in that... Step S2 also includes: One-dimensional spatial feature vectors are matched with each historical environmental spatial feature vector in the acoustic feature database one by one using cosine similarity. The vectors are sorted from largest to smallest according to their cosine similarity values. The historical environmental spatial feature vector at the top of the sort is used as the record corresponding to the current spatial pattern. Spatial acoustic wave reflectivity and medium transmittance are extracted synchronously from the same record as the historical environmental spatial feature vector at the top of the sort and written into the target acoustic mapping parameter set.

6. The method for controlling vibration noise in a beam fabrication plant based on big data as described in claim 1, characterized in that, Step S3 includes: In a three-dimensional virtual coordinate system that maintains the same coordinate semantics as the global three-dimensional coordinate system, real-time coordinate data is written into the virtual sound-generating node of the corresponding vibrating equipment, and the center coordinate of the air-transmitting three-dimensional spatial grid is used as the virtual receiving node. Virtual sound wave rays are established from each virtual sound-generating node to each virtual receiving node, and then continuous adjacent physical occlusion three-dimensional spatial grids are extracted along each virtual sound wave ray and merged to form a set of physical penetration segments.

7. The method for controlling vibration noise in a beam fabrication plant based on big data, as described in claim 6, is characterized in that... Step S3 also includes: For each set of physical penetration sections, the spatial acoustic reflectivity and medium transmittance in the target acoustic mapping parameter set are called to continuously calculate the interface reflection loss of the virtual acoustic ray at the incident and exit positions of the physical penetration section and the transmission attenuation inside the physical penetration section. Then, based on the initial nominal sound power and propagation distance, the single-source residual sound intensity of each virtual receiving node is obtained. All single-source residual sound intensities are accumulated and converted into residual sound pressure level values ​​to generate a dynamic sound field distribution prediction map.

8. The method for controlling vibration noise in a beam fabrication plant based on big data as described in claim 1, characterized in that, Step S4 includes: The first analytical quantity includes the cumulative value of the line-of-sight obstruction section of the direct sound wave; Based on the dynamic sound field distribution prediction map, spatial coordinates with residual sound pressure levels higher than the safety threshold are extracted as the set of noise exceeding coordinate nodes. The target vibrating equipment is locked according to the single-source residual sound intensity contribution of each vibrating equipment to each noise exceeding coordinate. A direct sound wave ray is established between the virtual sound-generating node corresponding to the target vibrating equipment and the target exceeding node. Then, the three-dimensional spatial mesh of intersecting entities that are continuously adjacent along the direct sound wave ray is connected and merged. The overall projection section after merging is calculated to obtain the cumulative value of the direct sound wave line-of-sight occlusion section.

9. A method for controlling vibration noise in a beam fabrication plant based on big data, as described in claim 8, is characterized in that... Step S4 also includes: The second analytical quantity includes the cumulative value of the line-of-sight obstruction section of the direct sound wave; The real-time drive current waveform data of the target vibrating equipment is collected synchronously. The frequency component with the highest amplitude in the design vibration response frequency band is extracted from the real-time drive current waveform data within the stable operation window as the frequency corresponding to the main frequency characteristic peak. The difference between the frequency corresponding to the main frequency characteristic peak and the standard rated vibration frequency of the target vibrating equipment is calculated to obtain the cumulative value of the direct sound wave line-of-sight obstruction section.

10. A method for controlling vibration noise in a beam fabrication plant based on big data, as described in claim 9, is characterized in that... Step S4 also includes: The accumulated values ​​of the direct line-of-sight obstruction section and the accumulated values ​​of the direct line-of-sight obstruction section are written into the same two-dimensional input vector in a fixed order. The two-dimensional input vector is then input into the pre-training judgment module, which is a support vector regression model with a radial basis function kernel. The support vector regression model outputs the intervention urgency coefficient. Based on the correspondence between the intervention urgency coefficient and the stepped tolerance threshold group, the system outputs a sound source control command signal to maintain the original operating frequency, a sound source control command signal to reduce the drive power supply frequency, or a sound source control command signal to cut off the drive power supply.

Images