Supercritical foaming material production management method for new energy battery protection

Abstract

The application discloses a supercritical foaming material production management method for new energy battery protection, and belongs to the technical field of supercritical foaming material production quality management, and specifically comprises the following steps: extracting a pressure relief stage sound wave signal historical segment of each space region, detecting a cell size and density, and establishing a batch acoustic feature database; extracting acoustic feature parameter combinations of sound wave signals of each space region and establishing a mapping relationship; extracting a deviation degree of acoustic features between each region in a high uniformity batch according to the mapping relationship to construct a deviation reference interval library; obtaining a current real-time sound wave signal and calculating a two-by-two deviation degree between each space region; when the deviation degree of any pair of space regions exceeds the reference interval library, the pair of regions is marked as a sampling inspection region, and a sampling inspection instruction containing a position identifier is generated. The application realizes directional marking of abnormal regions of the state of each space region in the foaming process, and significantly improves the detection ability and quality control efficiency of the cell structure uniformity in the batch.

Description

Technical Field

[0001] This invention relates to the field of supercritical foam material production quality management technology, specifically to a production management method for supercritical foam materials used for the protection of new energy batteries. Background Technology

[0002] The uniformity of the cell structure in supercritical foamed materials is a key factor determining the consistency of their buffering, energy absorption, heat insulation, and flame retardant properties, directly affecting the reliability of protective components for new energy power batteries. The basic process of supercritical foaming involves injecting supercritical fluid into a polymer matrix under high pressure to dissolve it and form a homogeneous system; then, rapid depressurization triggers bubble nucleation and growth; finally, cooling and solidification solidifies the cell structure. In industrial production, reactor volumes have increased from hundreds of liters to tens of thousands of liters. In large-sized reactors, the heat conduction path from the outer wall to the center is lengthened, and the difference in the distance the supercritical fluid diffuses from the injection port to different areas increases. This leads to significant non-uniformity in the temperature and pressure fields at different spatial locations within the reactor, resulting in deviations in cell size and density in foamed materials produced in different areas within the same batch.

[0003] To address the aforementioned spatial inhomogeneity issue, existing technologies have employed various methods for improvement. In terms of equipment structure, magnetically coupled stirring devices are used to enhance convection within the vessel, or multi-zone independent temperature control is employed to adjust heating power along the vessel wall, or the material is tumbled and flowed by rotating the vessel. Regarding process parameters, overall uniformity is optimized by adjusting foaming temperature, saturation pressure, and depressurization rate. However, these methods have not completely eliminated the differences in cell structure between different spatial regions. First, the stirring effect decreases with increasing distance from the stirring paddle; in high-viscosity polymer melts, the mixing effect still differs between the central region of the vessel and the region near the stirring paddle. Second, multi-zone independent temperature control is based on the vessel wall temperature, and there is a lag and deviation between the wall temperature and the actual temperature inside the polymer deep within the vessel. Third, process parameter adjustments are macroscopic adjustments for the entire vessel, making it difficult to provide targeted compensation for deviations in localized spatial regions.

[0004] Since the spatial non-uniformity cannot be completely eliminated, it is still necessary to confirm the quality of each batch of products through off-line sampling inspection. In the conventional method, after the completion of foaming, a sample is cut from the product, and the detected values of cell size and cell density are obtained through sample preparation and scanning electron microscope observation. The determination of sampling positions is usually based on preset fixed points, or sampling positions are selected on the product by random rules. However, in the above fixed sampling or random sampling methods, no correspondence is established between the spatial distribution of sampling positions on the product and the actual cell structure distribution in each region of the reaction kettle. When the cell size or cell density in a certain spatial region deviates from the overall average level of the batch, if the corresponding product position of this region is not covered by fixed sampling points and is not selected by random sampling, the deviation cannot be reflected by off-line detection. As a result, products with actual local cell structure abnormalities are still determined as qualified, and abnormal products flow into downstream processes, which may cause local failure of buffer energy absorption, heat insulation or flame retardancy performance of new energy power battery protection components during service, further leading to safety hazards such as thermal runaway spread or insufficient mechanical impact protection. SUMMARY OF THE INVENTION

[0005] An object of the present invention is to provide a production management method for supercritical foamed materials for new energy battery protection, and solve the following technical problem:

[0006] In actual production, the quality inspection of foamed products has long been in a "black box" or "semi-blind" state, which mainly relies on rough evaluation based on fixed or random off-line sampling points and batch average judgment, and it is difficult to achieve dynamic and refined detection based on the real cell structure distribution of each spatial region in the foaming still.

[0007] The object of the present invention can be achieved by the following technical solutions: A production management method for supercritical foamed materials for new energy battery protection, comprising the following steps: S1: based on historical production task execution data, extracting historical acoustic signal segments of a plurality of spatial regions in the same foaming kettle during the pressure relief stage, and the detected cell size values and detected cell density values of each spatial region, storing them in association with batch identifiers and spatial region identifiers, and establishing a batch acoustic feature database; S2: extracting an acoustic feature parameter combination from the historical acoustic signal segment of any spatial region in each batch in the batch acoustic feature database, and establishing a mapping relationship between the acoustic feature parameter combination and the detected cell size value and detected cell density value of the spatial region in the current batch; S3: according to the mapping relationship of each spatial region, marking batches in which the dispersion degree of cell size detection values and the dispersion degree of cell density detection values of each spatial region in the same batch are both lower than preset thresholds as high-uniformity batches, extracting the deviation degree between acoustic feature parameter combination sequences among various spatial regions in the high-uniformity batches, and constructing an inter-regional acoustic feature deviation reference interval library; S4: Obtain real-time acoustic signal segments of the current production batch during the depressurization stage, extract the real-time acoustic feature parameter combination sequence of each spatial region, and calculate the degree of deviation between the real-time acoustic feature parameter combination sequences of each spatial region within the current batch. S5: When the deviation between at least one pair of spatial regions exceeds the reference interval for acoustic feature deviation between regions, both pairs of spatial regions are taken as sampling areas, and a batch sampling decision instruction containing the location identifiers of each pair of spatial regions is generated.

[0008] As a further aspect of the present invention: in step S1, the specific process of dividing the plurality of spatial regions within the foaming vessel is as follows: M sensor layers are arranged axially on the inner wall of the reactor, and N acoustic wave sensors are arranged circumferentially on each sensor layer. The spatial positions of the acoustic wave sensors on the inner wall of the reactor are different from each other. With the geometric center of the internal space of the reactor as the origin, a three-dimensional spatial coordinate system is established with the axial direction as the X-axis and the two orthogonal directions in the radial plane as the Y-axis and Z-axis, and the spatial coordinates of each acoustic sensor in the three-dimensional spatial coordinate system are obtained. The internal space of the reactor is divided into M axial sections along the axis. Each of the M axial sections corresponds to one of the M sensor layers. The axial span of each axial section is the distance between two adjacent sensor layers. Each axial segment is divided into N circumferential sectors along the circumference. The N circumferential sectors correspond one-to-one with the N acoustic sensors in the sensor layer corresponding to the axial segment. The circumferential span of each circumferential sector is the arc length corresponding to the circumferential angle between two adjacent acoustic sensors. Each circumferential sector is divided into K radial layers, which are arranged sequentially from the inner wall of the reactor towards the central axis. The radial span of each radial layer is one-K times the radius of the reactor. A spatial unit defined by the same axial segment, the same circumferential sector, and the same radial layer is considered as a spatial region, resulting in a plurality of spatial regions.

[0009] As a further aspect of the present invention: the specific process of establishing the batch acoustic feature database in S1 is as follows: Obtain the production task execution data corresponding to several historical batches that have been produced in the supercritical foaming production system; For each historical batch, extract the historical segments of acoustic signals collected by acoustic sensors in multiple spatial regions during the pressure holding stage from the production task execution data corresponding to the historical batch, as well as the detection values ​​of bubble size and bubble density in each spatial region obtained by offline bubble structure detection in each spatial region of the historical batch. The acoustic signal historical segments corresponding to each historical batch are classified according to the spatial region from which they originate, thus obtaining a subset of acoustic signal historical segments corresponding to each historical batch in each spatial region. The batch identifier of each historical batch, the subset of historical acoustic signal segments corresponding to each historical batch in each spatial region, the bubble size detection value corresponding to the historical batch in that spatial region, and the bubble density detection value corresponding to the historical batch in that spatial region are associated to form a single batch acoustic feature record. The acoustic feature records of each batch corresponding to all historical batches are collected to obtain the batch acoustic feature database.

[0010] As a further aspect of the present invention: the specific process of classifying the historical segments of the acoustic signals corresponding to each historical batch according to their spatial regions of origin is as follows: Obtain the original acoustic signal records of all acoustic sensors during the pressure holding stage of a historical batch. The original acoustic signal records include acoustic signal segments collected by each acoustic sensor at each acquisition time, the spatial coordinates of each acoustic sensor, and the time stamp of each acquisition time. From acoustic signal segments of different acoustic sensors within a preset time window near the same acquisition time, feature waveforms with waveform similarity exceeding a preset similarity threshold are extracted, and the feature waveforms are determined to be excited by the same physical event. The spatial coordinates of the acoustic wave sensors that receive the characteristic waveform excited by the same physical event and the time stamp of each receiving the characteristic waveform are obtained. Based on the difference between the spatial coordinates of each acoustic wave sensor and the time stamp of each acoustic wave sensor receiving the characteristic waveform, an arrival time difference positioning equation is established. Solving the time-of-arrival positioning equation yields the coordinates of the location of the physical event within the reactor's internal containment space. The occurrence location coordinates are matched one by one with the boundary ranges of the plurality of spatial regions to determine the spatial region in which the occurrence location coordinates fall and the physical event is associated with that spatial region. By splicing together the acoustic signal fragments corresponding to all physical events associated with the same spatial region during the depressurization phase according to the acquisition time, a subset of the historical acoustic signal fragments of that historical batch in that spatial region is obtained.

[0011] As a further aspect of the present invention: the specific extraction process of the acoustic feature parameter combination in S2 is as follows: A spatial region is obtained from a batch of historical acoustic signal segments for signal separation processing, and the separated historical acoustic signal segments are divided into several equal-length windows. Perform a Fourier transform on the amplitude of the acoustic signal within each time window to obtain the spectral distribution corresponding to that time window. Extract the frequency component with the largest amplitude from the spectral distribution as the dominant frequency of that time window. Calculate the frequency value of the amplitude-weighted average of all frequency components in the spectral distribution as the spectral centroid of that time window. Combine the dominant frequency and spectral centroid of the same time window to form the frequency domain feature pair of that time window. Zero-crossing detection is performed on the acoustic signal amplitude within each time window. The number of times the acoustic signal amplitude crosses the zero amplitude line within that time window is counted as the zero-crossing count of that time window. The zero-crossing count of the same time window is used as the time-domain characteristic value of that time window. The frequency domain feature pairs and time domain feature values ​​of the same time window are combined to form the single-window acoustic feature vector of that time window. The single-window acoustic feature vectors of all time windows in the corresponding acoustic signal history segment of the spatial region are arranged in chronological order of the time windows to obtain the acoustic feature parameter combination.

[0012] As a further aspect of the present invention, the specific process of signal separation processing is as follows: When the high-pressure reactor is in an unloaded state without the injection of supercritical fluid and without the placement of polymer matrix, an unloaded depressurization operation with the same depressurization rate and the same initial depressurization pressure as a normal production batch is performed. Multiple acoustic wave sensors deployed on the inner wall of the high-pressure reactor collect acoustic wave signals throughout the entire unloaded depressurization operation process. The collected acoustic wave signals are used as the unloaded depressurization background acoustic wave signals corresponding to each acoustic wave sensor and stored. For any spatial region in a batch, the background acoustic signal of the acoustic sensor corresponding to the spatial region is obtained under no-load pressure relief. The acoustic amplitude of each sampling time in the original acoustic signal history segment is subtracted point by point from the acoustic amplitude of the background acoustic signal of the no-load pressure relief at the same sampling time to obtain the difference acoustic signal segment of the spatial region in the batch. The difference acoustic signal segment is input into a bandpass filter. The lower cutoff frequency of the bandpass filter is set to the lowest dominant frequency of the acoustic signal generated during the bubble merging process, and the upper cutoff frequency of the bandpass filter is set to the highest dominant frequency of the acoustic signal generated during the bubble nucleation process. The bandpass filter filters out signal components below the lower cutoff frequency and signal components above the upper cutoff frequency, retaining signal components between the lower and upper cutoff frequencies. The filtered difference acoustic signal segment is then used as the valid acoustic signal segment for that spatial region in that batch.

[0013] As a further aspect of the present invention: in step S3, the specific extraction process of the deviation is as follows: The acoustic feature parameter combination sequence of each spatial region within a high-uniformity batch is subjected to intra-batch normalization transformation. The single-window acoustic feature vector at the same time window position in the acoustic feature parameter combination sequence of each spatial region is subtracted from the mean of the single-window acoustic feature vectors of all spatial regions in the batch at the same time window position, and then divided by the standard deviation of the single-window acoustic feature vectors of all spatial regions in the batch at the same time window position to obtain the normalized single-window acoustic feature vector of each spatial region at each time window position. For any two spatial regions within a highly uniform batch, the difference between the standardized single-window acoustic feature vectors of the two spatial regions at the same time window position is squared to obtain a square value. The square values ​​at all time window positions are summed, and the square root of the sum is taken as the degree of deviation between the two spatial regions.

[0014] As a further aspect of the present invention: the specific construction process of the inter-regional acoustic feature deviation reference interval library in S3 is as follows: Extract all batches labeled as high-uniformity batches from the batch acoustic feature database to obtain a high-uniformity batch set. For each high uniformity batch in the set of high uniformity batches, obtain the acoustic feature parameter combination sequence of each spatial region within the high uniformity batch, and perform intra-batch normalization transformation on the acoustic feature parameter combination sequence of all spatial regions to obtain the normalized single-window acoustic feature vector of each spatial region at each time window position. For any two spatial regions within the high uniformity batch, the difference between the standardized single-window acoustic feature vectors of the two spatial regions at the same time window position is squared to obtain the square value. The square values ​​at all time window positions are summed, and the square root of the summation result is taken as the degree of pairwise deviation between the two spatial regions in the high uniformity batch. The minimum value among the pairwise deviations between all spatial regions within the high uniformity batch is taken as the lower limit of the deviation of the high uniformity batch, and the maximum value among the pairwise deviations between all spatial regions within the high uniformity batch is taken as the upper limit of the deviation of the high uniformity batch. The lower limit and the upper limit are combined to form the deviation range of the high uniformity batch. The deviation ranges of all high-uniformity batches in the high-uniformity batch set are aggregated to obtain the reference interval library of acoustic feature deviations between regions.

[0015] As a further aspect of the present invention: S5 further includes generating a batch sampling decision instruction to be executed according to a preset conventional sampling plan if the deviation between any two spatial regions does not exceed the reference interval library of acoustic feature deviations between regions.

[0016] The beneficial effects of this invention are: 1) This invention overcomes the limitation of existing technologies that rely solely on wall sensors or macroscopic parameters of the entire reactor to indirectly infer the internal state by constructing a global state perception capability based on the spatiotemporal localization of acoustic events. It is understood that during the depressurization stage of a supercritical fluid, the nucleation, growth, and merging of bubbles generate acoustic signals of specific frequencies. The propagation speed and path of these acoustic signals in the polymer melt are influenced by local temperature, viscosity, and the stage of bubble evolution. This invention, by arranging multiple acoustic sensors in an axial and circumferential array along the inner wall of the reactor and establishing a three-dimensional spatial coordinate system, utilizes the time difference between the characteristic waveforms excited by the same bubble physical event received by different sensors to establish an arrival time difference localization equation. This allows for the inverse calculation of the precise location coordinates of each bubble event, which are then assigned to a pre-divided spatial region consisting of axial segments, circumferential sectors, and radial layers. Thus, the discrete acoustic signal segments collected during the depressurization stage are reconstructed into a unique acoustic feature parameter combination sequence for each spatial region. This sequence is a dynamic response record of the bubble nucleation rate, growth rate, and merging degree in each region. In this way, in the complex high-temperature and high-pressure industrial reactor environment, non-invasive, spatially resolved direct perception of the foaming process at all spatial locations in the entire domain is achieved based on acoustic physics principles.

[0017] 2) This invention solves the problem of direct comparison between different batches due to the overall drift of acoustic characteristics caused by raw material fluctuations and environmental changes by establishing a uniformity evaluation benchmark based on intra-batch relative deviations. It is understood that highly uniform batches with highly uniform spatial distribution of bubble size and density in historical batches exhibit a stable "normal relative relationship" in the evolution of acoustic characteristics between different spatial regions within the reactor. This relative relationship is mainly determined by the inherent physical structure of the reactor, such as the fixed heating zone layout, agitator geometry, and injection port spatial position, and has cross-batch stability. This invention screens highly uniform batches, performs intra-batch standardization transformation on the acoustic characteristic parameter combination sequences of each spatial region, eliminates the influence of inter-batch overall drift, calculates the pairwise deviation between any two regions within the standardized space, and aggregates the pairwise deviation ranges of all highly uniform batches to form a reference interval library of inter-regional acoustic characteristic deviations. This reference interval library defines the normal fluctuation boundary of acoustic behavior differences between regions within the reactor under ideal homogeneous conditions. Its core logic is to use "relative relationship between regions" rather than "absolute value" as the evaluation criterion, so that the reference benchmark can remain effective across batch differences, providing a dynamic comparison basis based on historical high homogeneity samples for real-time evaluation of the homogeneity of the current production batch.

[0018] 3) This invention establishes a closed-loop sampling decision system from "global state perception" to "abnormal area directional calibration," overcoming the limitation of existing fixed or random sampling methods where the sampling location does not correspond to the actual spatial distribution of the bubble structure within the reactor. This invention acquires the real-time acoustic signal of the current production batch during the depressurization phase, extracts the real-time acoustic feature parameter combination sequence of each spatial region, performs the same batch-wide standardization transformation, calculates the pairwise deviation between each spatial region, and compares it against a reference interval library of inter-regional acoustic feature deviations. When the deviation of any pair of spatial regions exceeds the historical normal fluctuation boundary defined by the reference interval library, it indicates that at least one region in that pair exhibits a significantly deviated foaming behavior from the relative relationship that regions should maintain under highly uniform conditions, meaning that the local bubble structure of that region is likely abnormal. This invention simultaneously calibrates this pair of spatial regions as sampling areas and generates batch sampling decision instructions containing their location identifiers. This anomaly localization mechanism based on acoustic feature spatial deviation establishes a direct causal relationship between the offline sampling location and the actual foaming anomaly area inside the vessel revealed by the acoustic signal during the depressurization stage. It directs limited detection resources to the spatial location where local quality defects are most likely to exist, avoiding the risk of missed detection due to fixed or random sampling failing to cover the anomaly area. Attached Figure Description

[0019] The invention will now be further described with reference to the accompanying drawings.

[0020] Figure 1 This is a schematic diagram of the production management method for supercritical foaming materials used for the protection of new energy batteries according to the present invention. Detailed Implementation

[0021] 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.

[0022] Please see Figure 1 As shown, this invention relates to a production management method for supercritical foaming materials used for the protection of new energy batteries, comprising the following steps: S1: Based on historical production task execution data, extract historical segments of acoustic signals from multiple spatial regions within the same foaming kettle during the pressure relief stage, as well as the detection values ​​of bubble size and bubble density for each spatial region. Store these segments in association with batch identifiers and spatial region identifiers to establish a batch acoustic feature database. It is understandable that during the depressurization stage of supercritical fluids, the nucleation, growth, and coalescence of bubbles generate acoustic signals of specific frequencies. The propagation characteristics of these acoustic signals in the polymer melt are closely related to local temperature, viscosity, and the stage of bubble evolution. By arranging multiple acoustic sensors in an axial and circumferential array along the inner wall of the reactor, a three-dimensional spatial coordinate system is established, and the internal space of the reactor is divided into multiple spatial regions defined by axial segments, circumferential sectors, and radial layers. Each spatial region is then associated with a specific acoustic sensor combination. For each completed historical batch, the original acoustic signal records collected by all acoustic sensors during the depressurization stage are acquired. Using the time difference between the characteristic waveforms excited by the same bubble physical event received by different sensors, an arrival time difference localization equation is established to inversely deduce the precise location coordinates of each bubble event and assign it to the corresponding spatial region. The acoustic signal segments corresponding to all physical events associated with the same spatial region during the depressurization stage are spliced ​​together according to the acquisition time sequence to obtain a subset of historical acoustic signal segments for that historical batch in that spatial region. Simultaneously, after foaming, samples were cut from the corresponding product locations in each spatial region of the historical batch and examined using a scanning electron microscope to obtain the cell size and cell density values. The batch identifier of each historical batch, the subset of historical acoustic signal segments corresponding to that batch in each spatial region, and the corresponding cell size and cell density values ​​for that batch in that spatial region were correlated to form a single-batch acoustic feature record. All single-batch acoustic feature records corresponding to all historical batches were compiled to establish a batch acoustic feature database. This database establishes a strict spatiotemporal correspondence between discrete acoustic signals and the final cell structure quality indicators of each spatial region, providing data support for the establishment of subsequent mapping relationships.

[0023] S2: Extract acoustic feature parameter combinations from the historical segments of acoustic signals of any spatial region in each batch of the batch acoustic feature database, and establish a mapping relationship between the acoustic feature parameter combinations and the bubble size detection value and bubble density detection value of the spatial region in that batch; First, a historical segment of the acoustic signal corresponding to a batch in a spatial region is acquired and processed for signal separation. Specifically, with the high-pressure reactor in an unloaded state without supercritical fluid injection or polymer matrix, an unloaded depressurization operation with the same depressurization rate and initial pressure as a normal production batch is performed, and the background acoustic signal of the unloaded depressurization is collected and stored. For the original acoustic signal of this batch, its amplitude is subtracted from the background acoustic signal of the unloaded depressurization at each sampling moment to obtain the difference acoustic signal segment. Subsequently, the difference acoustic signal segment is input into a bandpass filter. The lower cutoff frequency of the bandpass filter is set to the lowest dominant frequency value of the acoustic signal generated by the bubble merging process, and the upper cutoff frequency is set to the highest dominant frequency value of the acoustic signal generated by the bubble nucleation process. This filters out interference signal components such as equipment mechanical vibration, fluid turbulence, and pipeline resonance, retaining the effective acoustic signal segment related to bubble behavior. Next, the separated effective acoustic signal segments are divided into several equal-length time windows. A Fourier transform is performed on the acoustic signal within each time window to obtain the spectral distribution. The frequency component with the largest amplitude is extracted as the dominant frequency. The frequency value obtained by weighting the amplitude of all frequency components is calculated as the spectral centroid. The dominant frequency and the spectral centroid are combined to form the frequency domain feature pair for that time window. Simultaneously, zero-crossing detection is performed on the acoustic signal within each time window, and the number of times it crosses the zero amplitude line is counted as the zero-crossing count, which is used as the time domain feature value for that time window. The frequency domain feature pairs and time domain feature values ​​of the same time window are combined to form a single-window acoustic feature vector. The single-window acoustic feature vectors of all time windows corresponding to that spatial region in that batch are arranged in chronological order to obtain the acoustic feature parameter combination sequence for that spatial region. A mapping relationship is established between the extracted acoustic feature parameter combination sequence and the bubble size and bubble density detection values ​​of that spatial region in that batch. This mapping relationship is established by traversing all spatial regions and all batches.

[0024] S3: Based on the mapping relationship of each spatial region, batches in the same batch whose dispersion of the bubble size detection value and the dispersion of the bubble density detection value in each spatial region are both lower than the preset threshold are marked as high uniformity batches. The deviation between the acoustic feature parameter combination sequences between each spatial region in the high uniformity batch is extracted, and a reference interval library of acoustic feature deviation between regions is constructed. It is understandable that the acoustic feature evolution trajectories between different spatial regions within the reactor in historically high-uniformity batches exhibit a stable "normal relative relationship." This relative relationship is primarily determined by the inherent physical structure of the reactor, such as the fixed heating zone layout, agitator geometry, and inlet spatial location, and thus possesses batch-wide stability. First, based on the mapping relationship of each spatial region, the dispersion of the bubble size detection values ​​and bubble density detection values ​​within each spatial region of the same batch is calculated. Batches with both dispersion values ​​below a preset threshold are labeled as high-uniformity batches. All batches labeled as high-uniformity batches are then extracted from the batch acoustic feature database to obtain a set of high-uniformity batches. For each highly uniform batch in the set of highly uniform batches, the acoustic feature parameter combination sequence of each spatial region within the highly uniform batch is obtained. Intra-batch normalization is then performed on the acoustic feature parameter combination sequences of all spatial regions. This involves subtracting the mean of the single-window acoustic feature vectors of all spatial regions within the batch at the same time window position from the single-window acoustic feature vector of the same spatial region in the combination sequence of each spatial region, and then dividing by the standard deviation of the single-window acoustic feature vectors of all spatial regions within the batch at the same time window position. This yields the standardized single-window acoustic feature vector of each spatial region at each time window position. Through intra-batch normalization, the overall drift of acoustic features caused by factors such as raw material fluctuations and environmental changes between different batches is eliminated, making the feature spaces of different batches comparable. For any two spatial regions within the highly uniform batch, the difference between the standardized single-window acoustic feature vectors of the two spatial regions at the same time window position is squared. The squared values ​​of all time window positions are summed, and the square root of the sum is taken as the pairwise deviation between the two spatial regions. The minimum deviation among all pairwise deviations within a highly uniform batch is taken as the lower limit of the deviation, and the maximum deviation as the upper limit; these are combined to form the deviation range of the highly uniform batch. The deviation ranges of all highly uniform batches in the set are then aggregated to obtain a reference interval library for inter-regional acoustic characteristic deviations. This reference interval library defines the normal fluctuation boundary of acoustic behavior differences between regions within the vessel under ideal homogeneous conditions.

[0025] S4: Obtain real-time acoustic signal segments of the current production batch during the depressurization stage, extract the real-time acoustic feature parameter combination sequence of each spatial region, and calculate the degree of deviation between the real-time acoustic feature parameter combination sequences of each spatial region within the current batch. During the depressurization operation of the current production batch, acoustic signal segments during the depressurization phase are collected in real time by acoustic sensors deployed on the inner wall of the reactor. For each spatial region, following the same signal separation, processing, and feature extraction procedure as in S2, a real-time acoustic feature parameter combination sequence is extracted from the corresponding real-time acoustic signal segment. Subsequently, the real-time acoustic feature parameter combination sequences of each spatial region within the current batch are subjected to the same intra-batch normalization transformation as in S3 to eliminate the influence of the overall level of the current batch, resulting in a standardized real-time single-window acoustic feature vector for each spatial region at each time window position. For any two spatial regions within the current batch, the pairwise deviation between the two spatial regions is calculated using the same deviation calculation method as in S3. By traversing all pairs of spatial regions within the current batch, a complete set of pairwise deviations for each spatial region in the current batch is obtained.

[0026] S5: When the deviation between at least one pair of spatial regions exceeds the reference interval for acoustic feature deviation between regions, both pairs of spatial regions are taken as sampling areas, and a batch sampling decision instruction containing the location identifiers of each pair of spatial regions is generated.

[0027] The pairwise deviations between spatial regions within the current batch calculated in S4 are compared one by one with the inter-regional acoustic characteristic deviation reference interval library constructed in S3. When the pairwise deviations of at least one pair of spatial regions exceed the historical normal fluctuation boundary defined in the reference interval library, it indicates that the foaming behavior of at least one region in that pair significantly deviates from the relative relationship that regions should maintain under a highly uniform state, meaning that there is a high probability of anomalies in the local cell structure of that region. In this case, the pair of spatial regions are simultaneously marked as sampling areas, and a batch sampling decision instruction containing the location identifiers of each pair of spatial regions is generated. This instruction is used to guide offline sampling personnel to cut samples from the foamed material product at the corresponding physical locations of the spatial regions for cell structure testing. When the pairwise deviations of all spatial regions do not exceed the inter-regional acoustic characteristic deviation reference interval library, it indicates that the relative acoustic characteristics between regions in the current batch are within the normal fluctuation range of historically highly uniform batches, and there is a high probability that the overall uniformity of the batch is good. In this case, a batch sampling decision instruction is generated to be executed according to the preset conventional sampling plan. This anomaly localization mechanism based on acoustic feature spatial deviation establishes a direct causal relationship between the offline sampling location and the actual foaming anomaly area inside the vessel revealed by the acoustic signal during the depressurization stage. It directs limited detection resources to the spatial location where local quality defects are most likely to exist, avoiding the risk of missed detection due to fixed or random sampling failing to cover the anomaly area.

[0028] It is worth noting that during the initial stage of actual production startup, the introduction of new products, or the transition period after significant adjustments to process parameters, the batch acoustic feature database may not have accumulated a sufficient number of historical batches that meet the high uniformity labeling conditions. This may result in the temporary inability to construct a reference interval library for acoustic feature deviations between regions, or insufficient sample size within the library to support reliable statistical comparisons. To address these situations, this method further includes the following emergency judgment logic: Real-time acoustic signal segments of the current production batch during the depressurization phase are acquired. Following the same signal separation and feature extraction process as S2, real-time acoustic feature parameter combination sequences for each spatial region are extracted. Using the same calculation method as S4, the pairwise deviation degrees between spatial regions within the current batch are obtained, forming a set of deviation degrees for the current batch. Subsequently, a statistical distribution analysis is performed on all pairwise deviation degrees within this set. The arithmetic mean of all deviation degrees in this set is calculated as the central level of the current batch deviation degree, and the standard deviation of all deviation degrees in this set is calculated as a statistical dispersion index to measure the dispersion of the set. The value obtained by adding a preset multiple of the statistical dispersion index to the central level of the current batch deviation degree is used as the relative deviation judgment boundary. The pairwise deviation values ​​of each pair of spatial regions in the current batch are compared with the relative deviation judgment limit one by one. When the pairwise deviation value of a pair of spatial regions is greater than the relative deviation judgment limit, it indicates that the mutual deviation of the pair of spatial regions within the current batch is significantly higher than the general level of difference between spatial regions in the batch.

[0029] Understandably, in the absence of established historical benchmarks, this emergency judgment logic shifts the evaluation criterion from "comparing the normal range of regional differences with historically high-uniformity batches" to "comparing the overall distribution level of differences between regions within the current batch." The underlying physical logic is that even if the overall foaming intensity of the current batch systematically differs from the historical optimal state due to fluctuations in raw material batches or deviations in process conditions, as long as the foaming behavior of each spatial region within the reactor is coordinated, the relative deviations between regions should be concentrated around a central level. Conversely, if the deviation between a pair of spatial regions significantly deviates from this distribution, it indicates that at least one region in that pair has decoupled its foaming behavior from the overall internal coordination of the current batch. This decoupling signifies a significantly increased probability of abnormal local cell structure in that region. Based on the above judgment results, spatial regions with deviation values ​​greater than the relative deviation judgment limit are simultaneously marked as sampling areas, and batch sampling decision instructions containing the location identifiers of each pair of spatial regions are generated. Thus, even in the absence of historical reference benchmarks, it is still possible to target and locate the spatial location most likely to have local pore structure anomalies based on the relative consistency analysis between spatial regions within the current batch, providing clear location guidance for offline sampling.

[0030] In a preferred embodiment of the present invention, the specific process of dividing the plurality of spatial regions within the foaming vessel in step S1 is as follows: First, M layers of sensors are arranged axially on the inner wall of the reactor. For example, the reactor is divided into three layers from bottom to top vertically: a lower sensor layer, a middle sensor layer, and an upper sensor layer. Each sensor layer has N acoustic sensors arranged circumferentially on the inner wall of the reactor. For example, eight acoustic sensors are evenly arranged circumferentially on each layer, with a central angle of 45 degrees between two adjacent acoustic sensors. In this way, a total of M multiplied by N acoustic sensors are arranged on the inner wall of the reactor. The installation positions of these acoustic sensors on the inner wall of the reactor are different, and each sensor occupies a unique combination of axial height and circumferential angle. The reason for adopting a layered array arrangement is that a single sensor can only sense acoustic signals within a limited range in its vicinity and cannot cover the entire space of a large reactor. Only by collecting data from multiple sensors from different directions at the same time can sufficient multi-view observation data be provided for subsequent sound source localization. Next, taking the geometric center of the internal space of the reactor as the origin, which is the point determined by the position of the center of the radial section at half the axial length of the reactor cavity, the axial direction of the reactor (from the bottom to the top) is taken as the X-axis, and two mutually perpendicular directions in the radial plane of the reactor are taken as the Y-axis and Z-axis, respectively, to establish a three-dimensional spatial coordinate system. In this coordinate system, the X-axis, Y-axis, and Z-axis coordinate values ​​of each acoustic sensor are obtained. These three coordinate values ​​are used together as the spatial coordinates of the acoustic sensor. The reason for establishing a spatial coordinate system and obtaining the precise coordinates of each sensor is that the accurate spatial position of each receiving sensor is required for subsequent sound source localization, so that the coordinates of the sound source can be deduced from the time difference of arrival. Then, the internal space of the reactor is divided into M axial sections along the axis. The number of axial sections is the same as the number of sensor layers. For example, if three layers of sensors are arranged as mentioned above, it is divided into three axial sections: the lower axial section, the middle axial section, and the upper axial section. The axial span of each axial section is equal to the axial distance between two adjacent sensor layers. The reason for this division is that the acoustic signals generated by acoustic events in each axial section are mainly captured by the sensor layers located near the upper and lower boundaries of that section. Therefore, by corresponding the axial sections with the sensor layers one by one, a mapping relationship between acoustic signals and axial positions can be established.Each axial segment is then divided into N circumferential sectors. The number of circumferential sectors is the same as the number of acoustic sensors deployed in the sensor layer corresponding to that axial segment. For example, each axial segment is divided into eight circumferential sectors. The circumferential span of each circumferential sector is equal to the arc length corresponding to the circumferential angle between two adjacent acoustic sensors, that is, the arc length corresponding to the 45-degree central angle. Each circumferential sector corresponds one-to-one with an acoustic sensor in that sensor layer. The reason for this division is that the directional information of the acoustic signal in the circumferential direction is mainly sensed by the sensor near the circumferential position. Corresponding the circumferential sectors to the circumferential positions of the sensors can establish a mapping relationship between the acoustic signal and the circumferential direction. Each circumferential sector is further divided into K radial layers. For example, from the inner wall of the reactor towards the central axis, it can be divided into three radial layers: the near-wall radial layer, the middle radial layer, and the central radial layer. The radial span of each radial layer is equal to the reactor radius divided by K, which is one-third of the radius. The reason for this division is that the acoustic signal attenuates when it propagates in the polymer melt. The farther away from the acoustic sensor, the weaker the energy of the acoustic signal when it reaches the sensor. Furthermore, the bubble nucleation and growth process differs between the region near the reactor wall and the region near the center of the reactor due to different temperature and pressure conditions. Therefore, it is necessary to divide different radial depths in the same circumferential orientation into independent regions to distinguish the foaming behavior at different radial positions. Finally, the spatial unit defined by the same axial segment, the same circumferential sector, and the same radial layer is considered as an independent spatial region. For example, the spatial unit located in the lower axial segment, the 45-degree circumferential sector, and the near-wall radial layer is an independent spatial region. Thus, the internal containment space of the entire reactor is divided into M multiplied by N multiplied by K spatial regions. Each spatial region has a clear spatial boundary and forms a spatial correspondence with a specific acoustic sensor combination, providing a spatial division basis for subsequently assigning the location of each bubble physical event to the corresponding spatial region.

[0031] The purpose of the above spatial region division method is to discretize the originally continuous internal space of the reactor into multiple independent regional units with clear spatial boundaries and closely related to the layout of the acoustic wave sensors, thereby establishing a strict and orderly correspondence between the acoustic wave signal acquisition location and the physical spatial location inside the reactor. The advantages of this division method are as follows: First, by corresponding the axial segments to the sensor layers, acoustic events at each axial height can be correctly assigned to the region at that height, avoiding confusion of bubble signals at different heights along the axial direction. Second, by corresponding the circumferential sectors to the circumferential positions of the sensors, bubble events at different circumferential orientations can be accurately distinguished, avoiding the loss or mismatch of circumferential information. Third, by dividing the radial layers, the bubble behavior near the reactor wall and near the center of the reactor can be monitored separately. This is because, in actual production, the radial temperature gradient of a large-sized reactor is one of the important reasons for differences in bubble structure; without radial layering, it is impossible to distinguish whether the abnormality is in the near-wall region or the central region. Overall, this spatial region division method decomposes the acoustic signal, which originally only reflected the overall average state inside the reactor, into independent acoustic response records of each local spatial region. Each spatial region corresponds to a unique historical segment of acoustic signal. This lays the spatial resolution foundation for subsequent steps to extract acoustic feature parameter combinations for each spatial region and establish their mapping relationship with cell size and density. It is precisely because of this refined spatial region division that the degree of deviation between each spatial region can be calculated in real time through acoustic signals during the depressurization stage of the current production batch. Based on the degree of deviation, it is possible to locate which one or more spatial regions have an abnormal situation of decoupling from the overall coordination relationship, thereby generating sampling decision instructions containing specific spatial region location markers. This allows offline sampling personnel to accurately find the physical location corresponding to the abnormal spatial region on the foamed material product for sampling and testing, realizing a complete technical chain from spatial analysis of acoustic signals inside the reactor to precise guidance of product sampling location.

[0032] In another preferred embodiment of the present invention, the specific process of establishing the batch acoustic feature database in S1 is as follows: First, obtain the production task execution data corresponding to several historical batches that have been completed in the supercritical foaming production system. These historical batches refer to batches that have completed the entire process from raw material input to foaming and molding and have undergone quality inspection before the current production task. For example, select fifty batches produced continuously in the past three months as a set of historical batches. Each batch has complete production task execution data, including the batch number, raw material grade and input amount, supercritical fluid injection pressure curve, depressurization rate curve, records of each temperature sensor, and raw acoustic signals collected by each acoustic sensor during the depressurization stage.

[0033] For each historical batch, historical segments of acoustic signals collected by acoustic sensors corresponding to multiple spatial regions during the depressurization phase are extracted from the production task execution data corresponding to that historical batch. It should be noted that the acoustic sensor corresponding to each spatial region does not refer to a single sensor, but rather to a group of sensors determined by the axial segment position and circumferential sector position of that spatial region within the reactor. This is because the acoustic waves generated by the bubble physical event occurring within a spatial region propagate in all directions and are received by multiple sensors. However, the sensors closest to the region and with the best orientation receive the strongest and clearest signals. Therefore, several acoustic sensors in the sensor layer corresponding to the axial segment of the spatial region, which are adjacent to the circumferential orientation of that spatial region, are considered as... The corresponding sensor group, for example, the spatial region located in the lower axial section and the circumferential sector at a 45-degree azimuth, includes three acoustic wave sensors located at 0 degrees, 45 degrees, and 90 degrees in the sensor layer of that axial section. The portion of the acoustic wave signal collected by these three sensors during the depressurization phase that is associated with the bubble event occurring in that spatial region is extracted as the historical segment of the acoustic wave signal of that spatial region. The extraction is based on the occurrence position coordinates of each bubble event obtained by solving the aforementioned time difference of arrival positioning equation. The acoustic wave signal segments corresponding to all bubble events whose occurrence position coordinates fall within the boundary range of that spatial region are spliced ​​together in chronological order of the event occurrence time to obtain the historical segment of the acoustic wave signal of that spatial region.

[0034] Simultaneously, for each historical batch, it is also necessary to obtain the cell size and cell density values ​​of each spatial region through offline cell structure detection. Specifically, after the foamed material products of the historical batch have completed cooling and shaping and are taken out of the reactor, according to the pre-established mapping relationship between the spatial region and the physical position of the product, for example, the spatial region located in the 45-degree circumferential sector of the lower axial section and the radial layer near the wall in the reactor corresponds to the sampling position at a specific azimuth angle at the bottom edge of the foamed material product. A sample of a specified size is cut from this sampling position. After sample preparation, the cell structure image is observed and photographed under a scanning electron microscope. The diameter of at least one hundred cells is measured from the image and its average value is calculated as the cell size detection value of the spatial region. At the same time, the number of cells per unit area is counted and converted into the number of cells per cubic centimeter as the cell density detection value of the spatial region. The above offline detection process is performed for each spatial region to obtain the cell size and cell density detection values ​​of the historical batch in all spatial regions.

[0035] Then, the historical segments of acoustic signals corresponding to each historical batch are classified according to the spatial region from which they originate. That is, all historical segments of acoustic signals from the same spatial region in the same historical batch are grouped together. For example, all acoustic signal segments from the 45-degree circumferential sector near the wall radial layer in the lower axial section of historical batch 023 are grouped together to obtain the subset of historical segments of acoustic signals corresponding to that historical batch in that spatial region. After traversing all spatial regions, the subset of historical segments of acoustic signals corresponding to that historical batch in each spatial region can be obtained. Next, the following four pieces of information are associated for each historical batch: batch identifier (e.g., batch number 023), subset of historical acoustic signal segments corresponding to each spatial region, cell size detection value corresponding to the historical batch in that spatial region, and cell density detection value corresponding to the historical batch in that spatial region. This forms a single-batch acoustic feature record. This record actually establishes a complete correspondence chain between the acoustic response during the depressurization stage and the final cell structure quality index within a specific spatial region of a historical batch. The batch identifier is used to distinguish different production batches, the subset of historical acoustic signal segments is a dynamic record of the foaming process in that spatial region, and the cell size detection value and cell density detection value are static representations of the foaming result in that spatial region. Finally, the acoustic feature records of each batch corresponding to all historical batches are collected. That is, the acoustic feature records of each batch of fifty historical batches in each spatial region are all stored in a unified database to obtain the batch acoustic feature database. The organization structure of this database can be understood as a two-dimensional table with spatial region as the first dimension and historical batch as the second dimension. Each cell in the table stores a subset of historical segments of acoustic signal in a specific historical batch in a specific spatial region and the corresponding bubble size detection value and bubble density detection value.

[0036] In another preferred embodiment of the present invention, the specific process of classifying the historical segments of the acoustic signals corresponding to each historical batch according to the spatial region from which they originated is as follows: First, we acquire the raw acoustic signal records collected by all acoustic sensors during the depressurization phase of a historical batch. These raw acoustic signal records contain acoustic signal segments recorded at every sampling moment during the entire depressurization process of the batch by all 24 acoustic sensors arranged on the inner wall of the reactor, for example, three layers with eight sensors per layer. The acoustic amplitude point is recorded every 10 microseconds at a sampling frequency of 100,000 times per second. Each acoustic signal segment contains waveform data of 1,000 consecutive sampling points, or 10 milliseconds. The raw acoustic signal records also contain the spatial coordinates of each acoustic sensor and the time stamp accurate to the microsecond level for each sampling moment. The reason for recording the spatial coordinates and time stamps is that sound source localization must rely on the spatial position difference between each receiving sensor and the time difference of signal arrival.

[0037] Next, feature waveforms with waveform morphology similarity exceeding a preset similarity threshold are extracted from acoustic signal segments within a preset time window near the same acquisition time of different acoustic sensors. For example, the 120th second after the start of depressurization is selected as the reference time, and a 10-millisecond time window is formed by extending 5 milliseconds forward and backward from this time. Within this time window, the acoustic signal segments acquired by all 24 acoustic sensors are examined. The degree of morphological similarity is evaluated by calculating the correlation coefficient between the waveforms of any two sensors within this time window. A correlation coefficient close to 1 indicates that the two waveforms are highly similar. When the waveform correlation coefficients of several sensors within the same time window all exceed the preset similarity threshold, such as 0.85, it is considered that these sensors have received feature waveforms excited by the same physical event. The feature waveform is then identified as an acoustic signal generated by the same bubble nucleation event, the same bubble growth event, or the same bubble merging event. The reason for using waveform morphology similarity instead of simply comparing amplitude is that although the amplitude of the acoustic waves excited by the same event decreases with distance during propagation, the waveform fluctuation structure remains relatively stable, while background noise or waveforms of different events do not have this consistency. Then, the spatial coordinates of each acoustic sensor that received the characteristic waveform excited by the same physical event, as well as the time stamp of each sensor's reception of the characteristic waveform, are obtained. For example, if four acoustic sensors simultaneously capture the characteristic waveform, the spatial coordinates of sensor 1 are 0.8m, -0.6m, 0.0m, and the reception time is 120 seconds and 3.450 milliseconds after the pressure relief begins; the spatial coordinates of sensor 2 are 0.8m, 0.6m, 0.0m, and the reception time is 120 seconds and 3.458 milliseconds; the spatial coordinates of sensor 3 are -0.8m, -0.6m, 0.0m, and the reception time is 120 seconds and 3.470 milliseconds; and the spatial coordinates of sensor 4 are -0.8m, 0.6m, 0.0 ... The distance is 1 meter and the reception time is 120 seconds and 3.478 milliseconds. Based on the spatial coordinates of these four sensors and the time difference between the time stamps of the characteristic waveforms received by each sensor, an arrival time difference positioning equation is established. The physical principle of this equation is that sound waves propagate at a relatively constant speed in a mixture of polymer matrix and supercritical fluid. The time it takes for a sound wave emitted from the same sound source to reach sensors at different spatial locations is proportional to the distance from the sensor to the sound source. Therefore, the time difference between any two sensors receiving the same signal is equal to the difference in distance from the two sensors to the sound source divided by the speed of sound wave propagation. By substituting the known spatial coordinates of each sensor and the measured time difference into the above relationship, a set of nonlinear equations with the spatial coordinates of the sound source as unknowns can be constructed.Next, the time difference of arrival positioning equation is solved to obtain the coordinates of the physical event's location within the reactor's internal containment space. The solution process involves iteratively approximating and continuously adjusting the guessed sound source location coordinates, substituting them into the equation to calculate the theoretical time difference, and comparing it with the measured time difference. When the error between the two is less than the preset tolerance, the guessed coordinates are the solution to the equation. Thus, the specific location of the bubble event in three-dimensional space is obtained, for example, coordinates of X-axis positive 0.35 meters, Y-axis negative 0.20 meters, and Z-axis positive 0.15 meters. Then, the coordinates of the location where the event occurred are matched one by one with the boundaries of a plurality of pre-divided spatial regions. For example, the X coordinate of 0.35 meters is compared with the boundaries of each axial segment and it is found that it falls into the middle axial segment. The Y coordinate of -0.20 meters and the Z coordinate of 0.15 meters are compared with the boundaries of each circumferential sector within the axial segment and it is found that they fall into the 90-degree azimuth circumferential sector. At the same time, the distance from the coordinate point to the central axis is calculated and compared with the boundary of the radial layer and it is found that it falls into the middle radial layer. Thus, it is determined that the physical event belongs to the spatial region jointly defined by the middle axial segment, the 90-degree azimuth circumferential sector and the middle radial layer. Finally, the acoustic signal segments corresponding to all physical events associated with the same spatial region during the depressurization phase are spliced ​​together according to the order of acquisition time. For example, during the period from the 120th to the 125th second of the depressurization phase, there are 327 bubble nucleation events, 189 bubble growth events, and 43 bubble merging events associated with the above-mentioned spatial region. The acoustic signal segments corresponding to these events are spliced ​​together end to end according to the order of occurrence time to form a continuous acoustic signal segment, thus obtaining a subset of the historical acoustic signal segments of this historical batch in this spatial region. After traversing all spatial regions, the spatial region classification of the acoustic signal of this historical batch is completed.

[0038] In another preferred embodiment of the present invention, the specific extraction process of the acoustic feature parameter combination in S2 is as follows: The effective acoustic signal segments after separation and processing are divided into several equal-length windows along the time axis. For example, the effective acoustic signal segments of the entire depressurization stage are divided into time windows of 10 milliseconds each. If the entire depressurization stage lasts for 30 seconds, a total of 3,000 equal-length windows are divided. Each time window contains 1,000 sampling points. Since the sampling frequency is 100,000 times per second, 10 milliseconds correspond to 1,000 sampling points. The reason for using equal-length windows is that the acoustic characteristics of bubble nucleation and growth are relatively stable in a short time but have an evolution trend over a longer time. Therefore, it is necessary to discretize the continuous signal into a series of short-time segments through time windows in order to extract feature parameters window by window. Then, a Fourier transform is performed on the acoustic signal amplitude within each time window to obtain the corresponding spectral distribution. The physical principle of the Fourier transform is that any complex periodic or non-periodic waveform can be decomposed into a series of sine waves with different frequencies and amplitudes. By transforming the time-domain acoustic wave amplitude signal that fluctuates with time to the frequency domain through the Fourier transform, the energy distribution of the acoustic signal at each frequency component within the time window, i.e., the spectral distribution, is obtained. The frequency component with the largest amplitude is extracted from the spectral distribution as the dominant frequency of the time window. The dominant frequency reflects the most important oscillation frequency of the bubble acoustic activity within the time window. For example, the bubble within a certain time window... Nucleation events occur frequently, and their dominant frequency may be around 22,000 Hz. Simultaneously, the frequency value of the amplitude-weighted average of all frequency components in the spectral distribution is calculated as the centroid of the spectrum for that time window. The centroid is calculated by multiplying the amplitude of each frequency component by the frequency value, summing the results, and then dividing by the sum of the amplitudes of all frequency components. The centroid reflects the position of the center of gravity of the acoustic signal energy on the frequency axis within that time window. For example, when the energy of high-frequency components increases, the centroid shifts towards higher frequencies; when the energy of low-frequency components increases, the centroid shifts towards lower frequencies. The dominant frequency and centroid of the same time window are combined to form the frequency domain characteristic pair of that time window.

[0039] Simultaneously, zero-crossing detection is performed on the acoustic signal amplitude within each time window. The number of times the acoustic signal amplitude crosses the zero amplitude line within that time window is taken as the zero-crossing count for that time window. The principle of zero-crossing detection is that the complexity and frequency of the acoustic signal are closely related to the frequency of the waveform crossing the zero axis. When bubble nucleation events are frequent and bubble oscillation modes are rich, the acoustic waveform fluctuates violently and the number of times it crosses the zero axis increases accordingly. Conversely, when bubble activity is slow, the number of times it crosses the zero axis decreases. The number of zero-crossings in the same time window is taken as the time-domain characteristic value of that time window. Finally, the frequency domain feature pairs (i.e., the dominant frequency and the spectral centroid) of the same time window are combined with the time domain feature value (i.e., the number of zero crossings) to form the single-window acoustic feature vector of that time window. For example, the single-window acoustic feature vector of a certain time window contains three values: a dominant frequency of 22,000 Hz, a spectral centroid of 19,500 Hz, and 85 zero crossings. The single-window acoustic feature vectors of all time windows in the effective acoustic signal segment corresponding to the spatial region in this batch are arranged in the order of the time windows to form a time sequence vector. This time sequence vector is the acoustic feature parameter combination of the spatial region in this batch. For example, 3,000 time windows correspond to 3,000 single-window acoustic feature vectors, which are arranged in the order of the first time window, the second time window, and so on up to the third 3,000 time windows to form the acoustic feature parameter combination sequence of the spatial region.

[0040] The dominant frequency reflects the type of bubble behavior that is dominant within the time window, while the centroid of the spectrum reflects the trend of the overall frequency distribution's centroid change. By extracting time-domain feature parameters through zero-crossing detection, the intensity of bubble activity can be assessed from the perspective of signal complexity. A higher number of zero-crossings indicates a more complex acoustic waveform, corresponding to denser bubble nucleation or more intense bubble oscillation. By dividing the continuous signal into equal-length time windows and extracting feature vectors for each window, and then arranging the feature vectors of all time windows into a sequence in chronological order, the original one-dimensional time-domain waveform is transformed into a multi-dimensional feature sequence. This not only preserves the dynamic information of the bubble formation process over time but also significantly compresses the data volume, facilitating subsequent cross-regional and cross-batch comparative analysis. Transforming the acoustic response of each spatial region during the depressurization phase into a structured sequence of acoustic characteristic parameters enables subsequent steps to quantify the consistency of foaming behavior in each region by calculating the degree of deviation between the acoustic characteristic parameter combination sequences of different spatial regions. This allows for the identification of abnormal regions decoupled from the overall coordination relationship in the current production batch by comparing whether the pairwise deviation between each spatial region exceeds the normal deviation range established by historically high-uniformity batches, thus providing precise spatial location guidance for targeted sampling inspection.

[0041] In another preferred embodiment of the present invention, the specific process of signal separation processing is as follows: First, a historical acoustic signal segment corresponding to a specific spatial region in a batch is acquired and processed for signal separation. The purpose of this signal separation is to separate the effective acoustic signals generated during bubble nucleation, growth, and merging from background interference signals such as equipment mechanical vibration, fluid turbulence noise, and pipeline resonance. Specifically, an unloaded depressurization operation with the exact same depressurization rate and initial pressure as a normal production batch is performed in a high-pressure reactor under unloaded conditions without supercritical fluid injection or polymer matrix. The acoustic signals of the entire unloaded depressurization operation are collected by 24 acoustic sensors deployed on the inner wall of the reactor, and the collected acoustic signals are stored as the unloaded depressurization background acoustic signals corresponding to each acoustic sensor. For the original historical acoustic signal segment corresponding to this spatial region in that batch, the unloaded depressurization background acoustic signal of the corresponding acoustic sensor in that spatial region is acquired. The acoustic amplitude at each sampling moment in the original historical acoustic signal segment is subtracted point by point from the acoustic amplitude at the same sampling moment in the unloaded depressurization background acoustic signal. The difference acoustic signal segment is obtained. The reason why point-by-point subtraction can remove background interference is that in the no-load state there is no polymer matrix and therefore no bubble behavior. The acoustic signal collected at this time comes purely from the mechanical vibration of the equipment and the noise of fluid flow. In normal production batches, these background interferences also exist and their waveform characteristics are relatively stable. Therefore, subtracting the no-load background signal from the original signal of the normal batch can retain the net acoustic signal generated by bubble behavior. Then, the difference acoustic signal segment is input into a bandpass filter. The lower cutoff frequency of the bandpass filter is set to the lowest dominant frequency value of the acoustic signal generated by the bubble merging process, such as 8000 Hz, and the upper cutoff frequency is set to the highest dominant frequency value of the acoustic signal generated by the bubble nucleation process, such as 35000 Hz. The bandpass filter filters out low-frequency interference signal components below 8000 Hz and high-frequency interference signal components above 35000 Hz, retaining the signal components between 8000 Hz and 35000 Hz. The filtered difference acoustic signal segment is the effective acoustic signal segment of this spatial region in this batch.

[0042] By effectively eliminating background interference signals unrelated to bubble behavior, such as equipment mechanical vibration, fluid turbulence, and pipeline resonance, the extracted acoustic feature parameter combination truly reflects the physical information of the bubble nucleation, growth, and merging process itself, rather than mixed information contaminated by interference signals. By extracting frequency domain feature parameters through Fourier transform, it is possible to distinguish different bubble behavior modes from a frequency perspective, because the sound wave frequency generated by bubble nucleation is relatively high, usually concentrated between 20,000 Hz and 35,000 Hz, while the sound wave frequency generated by bubble merging is relatively low, usually concentrated between 8,000 Hz and 15,000 Hz.

[0043] In another preferred embodiment of the present invention, the specific extraction process of the deviation in S3 is as follows: The acoustic feature parameter combination sequence of each spatial region within a high-uniformity batch is subjected to intra-batch normalization transformation. The single-window acoustic feature vector at the same time window position in the acoustic feature parameter combination sequence of each spatial region is subtracted from the mean of the single-window acoustic feature vectors of all spatial regions in the batch at the same time window position, and then divided by the standard deviation of the single-window acoustic feature vectors of all spatial regions in the batch at the same time window position to obtain the normalized single-window acoustic feature vector of each spatial region at each time window position. For any two spatial regions within a highly uniform batch, the difference between the standardized single-window acoustic feature vectors of the two spatial regions at the same time window position is squared, and the squared values ​​at all time window positions are summed. The square root of the summation result is taken as the degree of deviation between the two spatial regions.

[0044] In another preferred embodiment of the present invention, the specific construction process of the inter-regional acoustic feature deviation reference interval library in step S3 is as follows: First, all batches marked as high-uniformity batches are extracted from the batch acoustic feature database to obtain a high-uniformity batch set. For example, among the 50 historical batches accumulated, 15 batches are marked as high-uniformity batches after the dispersion determination in step S3. These 15 high-uniformity batches constitute the high-uniformity batch set. Each high-uniformity batch contains complete acoustic feature parameter combination sequences for each of the 24 spatial regions, as well as corresponding bubble size and bubble density detection values. Then, for each high-uniformity batch in the high-uniformity batch set, the deviation range extraction operation is performed one by one. Taking the first high-uniformity batch as an example, the specific process is explained. First, the acoustic feature parameter combination sequences of all 24 spatial regions in the high-uniformity batch are obtained. The acoustic feature parameter combination sequence of each spatial region is composed of 3000 time windows of single-window acoustic feature vectors arranged in chronological order. Each single-window acoustic feature vector contains three feature values: the main frequency value, the spectral centroid value, and the number of zero crossings.

[0045] Next, batch-level normalization was performed on the acoustic feature parameter combination sequences of all 24 spatial regions. Specifically, for the first time window, the dominant frequency values ​​in the single-window acoustic feature vectors of all 24 spatial regions were extracted, and the arithmetic mean and standard deviation of these 24 dominant frequency values ​​were calculated. Similarly, the mean and standard deviation of the 24 spectral centroid values ​​and the mean and standard deviation of the 24 zero-crossing times were calculated. Then, the dominant frequency value of each spatial region in that time window was subtracted from the mean dominant frequency value of that time window, and then divided by the standard deviation of the dominant frequency value to obtain the normalized dominant frequency value. Finally, the spectral centroid value was subtracted from the mean spectral centroid value and then divided by the standard deviation of the spectral centroid value to obtain the normalized dominant frequency value. The standardized spectral centroid value is obtained by subtracting the mean of the zero-crossing count from the zero-crossing count value and then dividing by the standard deviation of the zero-crossing count. These three standardized values ​​together constitute the standardized single-window acoustic feature vector of the spatial region at the time window position. Subsequently, the above standardization operation is repeated for the second time window, the third time window, and up to the 3000th time window. For each time window, the mean and standard deviation of each spatial region are calculated based on the feature values ​​of all 24 spatial regions at the time window position, and the standardization transformation of each spatial region is completed. Finally, the standardized single-window acoustic feature vector sequence of each of the 24 spatial regions in the high uniformity batch at the 3000th time window position is obtained.

[0046] The reason for calculating the mean and standard deviation separately for each time window and performing independent standardization instead of using a uniform mean and standard deviation for the entire sequence is that acoustic characteristics exhibit significant evolutionary trends at different time points during the decompression phase. For example, in the early stage of decompression, bubble nucleation is dense, the dominant frequency is high, and the number of zero crossings is large. In the later stage of decompression, bubble growth stabilizes, the dominant frequency decreases, and the number of zero crossings decreases. If a uniform standardization is used, this dynamic evolutionary information in the time dimension will be smoothed out. However, time window-by-time standardization preserves the differences in the relative relationships between spatial regions within each time window, so that subsequent deviation calculations can accurately reflect the relative behavioral deviations of each spatial region at the same moment.

[0047] After completing the intra-batch normalization transformation, for any two spatial regions within the highly uniform batch, calculate the pairwise deviation between them. For example, select spatial regions 1 and 2, and compare the normalized single-window acoustic feature vector of spatial region 1 in the first time window with the normalized single-window acoustic feature vector of spatial region 2 in the first time window. This vector contains three components: normalized dominant frequency value, normalized spectral centroid value, and normalized zero-crossing number value. Calculate the normalized dominant frequency value of spatial region 1 minus the normalized dominant frequency value of spatial region 2 to obtain the dominant frequency difference, and take the square of this difference. Calculate the normalized spectral centroid value of spatial region 1 minus the normalized zero-crossing number value of spatial region 2. The difference between the centroids of the normalized spectrum of the spatial region is obtained by taking the square of the difference. The difference between the zero-crossing times of the normalized spectrum of the first spatial region and the zero-crossing times of the normalized spectrum of the second spatial region is obtained by taking the square of the difference. The three squared values ​​are added together to obtain the single-window squared deviation between the two spatial regions at the time window position. The single-window squared deviation of each time window from the first time window to the 3000th time window is accumulated in the same way to obtain a cumulative squared deviation. The square root of the cumulative squared deviation is the pairwise deviation between the first spatial region and the second spatial region in the high uniformity batch.

[0048] The degree of deviation between the two pairs reflects the comprehensive difference in the acoustic feature evolution trajectory of the two spatial regions throughout the entire depressurization stage. Since the calculation process uses the method of summing the squared differences of time windows and then taking the square root, the deviation at any time is accumulated and included in the final result, thus ensuring that the degree of deviation can fully characterize the synchronicity of the behavior of the two spatial regions throughout the entire foaming process.

[0049] Following the above method, all possible combinations of two spatial regions within the highly uniform batch are traversed. There are 276 pairs of spatial regions across the 24 regions, resulting in 276 pairwise deviation values. The minimum of these 276 pairwise deviation values ​​is then used as the lower limit of the deviation for the highly uniform batch. For example, if the two spatial regions with the most similar acoustic behavior within the batch have a deviation value of 2.8, then 2.8 is the lower limit of the deviation for this highly uniform batch. Simultaneously, the maximum of these 276 pairwise deviation values ​​is used as the upper limit of the deviation for this highly uniform batch. For example, if the two spatial regions with the greatest difference in acoustic behavior within the batch have a deviation value of 6.5, then 6.5 is the upper limit of the deviation for this highly uniform batch. Combining the lower limit of 2.8 and the upper limit of 6.5 forms the deviation range of the highly uniform batch, from 2.8 to 6.5. The physical meaning of this deviation range lies in the fact that it defines the range of normal acoustic behavior differences between different spatial regions under the ideal state of highly uniform spatial distribution of pore size and pore density, determined by the inherent physical structure of the reactor.

[0050] Following the exact same method described above, the deviation range extraction operation was performed on the 2nd, 3rd, and up to the 15th high-uniformity batches in the high-uniformity batch set. Each high-uniformity batch yielded its own corresponding lower and upper deviation limits. For example, the deviation range for the 2nd high-uniformity batch was 2.5 to 6.0, the 3rd high-uniformity batch's deviation range was 3.0 to 6.8, and so on, resulting in 15 deviation ranges. Finally, the deviation ranges of all 15 high-uniformity batches in the high-uniformity batch set were aggregated, that is, the 15 lower and 15 upper deviation limits were combined to form a reference interval library for inter-regional acoustic feature deviations. This reference interval library is essentially a reference set composed of multiple normal deviation ranges, covering the deviation fluctuation range from 2.5 to 6.8. The reason for not further compressing the deviation range of all high-uniformity batches into a single interval, but instead retaining multiple intervals, is that due to slight differences in process conditions, the normal deviation levels between the internal spatial regions of different high-uniformity batches may vary slightly. Retaining multiple intervals can more completely reflect the diversity of normal states and avoid misjudging normal fluctuations as abnormalities due to excessive compression.

[0051] By extracting multiple highly uniform batches instead of relying solely on a single batch to construct a reference interval library, it is possible to effectively avoid deviations in the reference benchmark caused by accidental factors in a particular batch. For example, although the cell structure uniformity of a highly uniform batch meets the requirements, the acoustic signal during its decompression stage may be subject to accidental external interference, causing the deviation between different regions within that batch to be abnormally large or small. If only the deviation range of that single batch is used as the reference benchmark, it may lead to misjudgments in subsequent evaluations. However, by aggregating the deviation ranges of multiple highly uniform batches and taking the minimum value among all lower limits and the maximum value among all upper limits as the overall reference interval library, it can cover the entire normal fluctuation range of the deviation between different regions under historical highly uniform conditions, making the reference benchmark more robust and reliable.

[0052] By eliminating the overall differences in acoustic characteristics caused by fluctuations in the melt index of raw materials, changes in ambient temperature, and equipment status drift between different high-uniformity batches through intra-batch standardization transformation, the deviation calculated from different batches is comparable across batches. This is because the standardization transformation uniformly converts the acoustic characteristic parameters of each batch to a relative scale with the overall level of the batch itself as the reference benchmark. The mean and standard deviation of each characteristic dimension are normalized. Therefore, the deviation calculated from different batches in the standardized characteristic space reflects the relative deviation of each spatial region relative to the overall coordination level within the batch, rather than the absolute numerical difference. This makes it statistically reasonable and effective to aggregate the deviation ranges of multiple high-uniformity batches together to construct a reference interval library. The establishment of a reference interval library for acoustic feature deviations between regions through the above process enables the comparison of the pairwise deviations between spatial regions within the current batch with the normal fluctuation range of the deviations between regions under historical high uniformity conditions when processing the current production batch. When the pairwise deviations of a pair of spatial regions exceed the normal fluctuation boundary defined by the reference interval library, it indicates that at least one region in the pair of spatial regions has significantly deviated from the relative coordination relationship that regions should maintain under high uniformity conditions. This allows the pair of spatial regions to be simultaneously identified as sampling areas and a sampling decision instruction containing its location identifier to be generated, realizing a complete decision chain from historical high uniformity sample statistics to the location of abnormal areas in the current batch.

[0053] In another preferred embodiment of the present invention, step S5 further includes generating a batch sampling decision instruction to be executed according to a preset conventional sampling plan if the deviation between any two spatial regions does not exceed the reference interval library of acoustic feature deviations between regions.

[0054] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A production management method for supercritical foamed materials used for the protection of new energy batteries, characterized in that, Includes the following steps: S1: Based on historical production task execution data, extract historical segments of acoustic signals from multiple spatial regions within the same foaming kettle during the pressure relief stage, as well as the detection values ​​of bubble size and bubble density for each spatial region. Store these segments in association with batch identifiers and spatial region identifiers to establish a batch acoustic feature database. S2: Extract acoustic feature parameter combinations from the historical segments of acoustic signals of any spatial region in each batch of the batch acoustic feature database, and establish a mapping relationship between the acoustic feature parameter combinations and the bubble size detection value and bubble density detection value of the spatial region in that batch; S3: Based on the mapping relationship of each spatial region, batches in the same batch whose dispersion of the bubble size detection value and the dispersion of the bubble density detection value in each spatial region are both lower than the preset threshold are marked as high uniformity batches. The deviation between the acoustic feature parameter combination sequences between each spatial region in the high uniformity batch is extracted, and a reference interval library of acoustic feature deviation between regions is constructed. The specific construction process of the inter-regional acoustic feature deviation reference interval library is as follows: Extract all batches labeled as high-uniformity batches from the batch acoustic feature database to obtain a high-uniformity batch set. For each high uniformity batch in the set of high uniformity batches, obtain the acoustic feature parameter combination sequence of each spatial region within the high uniformity batch, and perform intra-batch normalization transformation on the acoustic feature parameter combination sequence of all spatial regions to obtain the normalized single-window acoustic feature vector of each spatial region at each time window position. For any two spatial regions within the high uniformity batch, the difference between the standardized single-window acoustic feature vectors of the two spatial regions at the same time window position is squared to obtain the square value. The square values ​​at all time window positions are summed, and the square root of the summation result is taken as the degree of pairwise deviation between the two spatial regions in the high uniformity batch. The minimum value among the pairwise deviations between all spatial regions within the high uniformity batch is taken as the lower limit of the deviation of the high uniformity batch, and the maximum value among the pairwise deviations between all spatial regions within the high uniformity batch is taken as the upper limit of the deviation of the high uniformity batch. The lower limit and the upper limit are combined to form the deviation range of the high uniformity batch. The deviation ranges of all high-uniformity batches in the high-uniformity batch set are collected, and the minimum value among all lower limits and the maximum value among all upper limits are taken to form the deviation fluctuation space, thus obtaining the reference interval library of acoustic feature deviations between regions. S4: Obtain real-time acoustic signal segments of the current production batch during the depressurization stage, extract the real-time acoustic feature parameter combination sequence of each spatial region, and calculate the degree of deviation between the real-time acoustic feature parameter combination sequences of each spatial region within the current batch. S5: When the deviation between at least one pair of spatial regions exceeds the reference interval for acoustic feature deviation between regions, both pairs of spatial regions are taken as sampling areas, and a batch sampling decision instruction containing the location identifiers of each pair of spatial regions is generated.

2. The production management method for supercritical foamed materials used for the protection of new energy batteries according to claim 1, characterized in that, In step S1, the specific process of dividing the plurality of spatial regions within the foaming vessel is as follows: M sensor layers are arranged axially on the inner wall of the foaming kettle, and N acoustic wave sensors are arranged circumferentially on each sensor layer. The spatial positions of the acoustic wave sensors on the inner wall of the foaming kettle are different from each other. With the geometric center of the internal space of the foaming vessel as the origin, a three-dimensional spatial coordinate system is established with the axial direction as the X-axis and the two orthogonal directions in the radial plane as the Y-axis and Z-axis, and the spatial coordinates of each acoustic sensor in the three-dimensional spatial coordinate system are obtained. The internal space of the foaming vessel is divided into M axial sections along the axis. Each of the M axial sections corresponds to one of the M sensor layers. The axial span of each axial section is the distance between two adjacent sensor layers. Each axial segment is divided into N circumferential sectors along the circumference. The N circumferential sectors correspond one-to-one with the N acoustic sensors in the sensor layer corresponding to the axial segment. The circumferential span of each circumferential sector is the arc length corresponding to the circumferential angle between two adjacent acoustic sensors. Each circumferential sector is divided into K radial layers. The K radial layers are arranged sequentially from the inner wall of the foaming vessel towards the central axis. The radial span of each radial layer is one-Kth of the radius of the foaming vessel. A spatial unit defined by the same axial segment, the same circumferential sector, and the same radial layer is considered as a spatial region, resulting in a plurality of spatial regions.

3. The production management method for supercritical foamed materials used for the protection of new energy batteries according to claim 1, characterized in that, In S1, the specific process of establishing the batch acoustic feature database is as follows: Obtain the production task execution data corresponding to several historical batches that have been produced in the supercritical foaming production system; For each historical batch, extract the historical segments of acoustic signals collected by acoustic sensors in multiple spatial regions during the depressurization phase from the production task execution data corresponding to the historical batch, as well as the detection values ​​of bubble size and bubble density of each spatial region obtained by offline bubble structure detection in each spatial region of the historical batch. The acoustic signal historical segments corresponding to each historical batch are classified according to the spatial region from which they originate, thus obtaining a subset of acoustic signal historical segments corresponding to each historical batch in each spatial region. The batch identifier of each historical batch, the subset of historical acoustic signal segments corresponding to each historical batch in each spatial region, the bubble size detection value corresponding to the historical batch in that spatial region, and the bubble density detection value corresponding to the historical batch in that spatial region are associated to form a single batch acoustic feature record. The acoustic feature records of each batch corresponding to all historical batches are collected to obtain the batch acoustic feature database.

4. The production management method for supercritical foamed materials used for the protection of new energy batteries according to claim 3, characterized in that, The specific process of classifying the historical segments of acoustic signals corresponding to each historical batch according to their spatial regions of origin is as follows: Obtain the original acoustic signal records of all acoustic sensors in a historical batch during the pressure relief phase. The original acoustic signal records include acoustic signal segments collected by each acoustic sensor at each acquisition time, the spatial coordinates of each acoustic sensor, and the time stamp of each acquisition time. From acoustic signal segments of different acoustic sensors within a preset time window near the same acquisition time, feature waveforms with waveform similarity exceeding a preset similarity threshold are extracted, and the feature waveforms are determined to be excited by the same physical event. The spatial coordinates of the acoustic wave sensors that receive the characteristic waveform excited by the same physical event and the time stamp of each receiving the characteristic waveform are obtained. Based on the difference between the spatial coordinates of each acoustic wave sensor and the time stamp of each acoustic wave sensor receiving the characteristic waveform, an arrival time difference positioning equation is established. Solve the arrival time difference positioning equation to obtain the location coordinates of the physical event within the containment space of the foaming kettle; The occurrence location coordinates are matched one by one with the boundary ranges of the plurality of spatial regions to determine the spatial region in which the occurrence location coordinates fall and the physical event is associated with that spatial region. By splicing together the acoustic signal fragments corresponding to all physical events associated with the same spatial region during the depressurization phase according to the order of acquisition time, a subset of historical acoustic signal fragments for that historical batch in that spatial region is obtained.

5. The production management method for supercritical foaming materials used for the protection of new energy batteries according to claim 1, characterized in that, In S2, the specific extraction process of the acoustic feature parameter combination is as follows: A spatial region is obtained from a batch of historical acoustic signal segments for signal separation processing, and the separated historical acoustic signal segments are divided into several equal-length windows. Perform a Fourier transform on the amplitude of the acoustic signal within each time window to obtain the spectral distribution corresponding to that time window. Extract the frequency component with the largest amplitude from the spectral distribution as the dominant frequency of that time window. Calculate the frequency value of the amplitude-weighted average of all frequency components in the spectral distribution as the spectral centroid of that time window. Combine the dominant frequency and spectral centroid of the same time window to form the frequency domain feature pair of that time window. Zero-crossing detection is performed on the acoustic signal amplitude within each time window. The number of times the acoustic signal amplitude crosses the zero amplitude line within that time window is counted as the zero-crossing count of that time window. The zero-crossing count of the same time window is used as the time-domain characteristic value of that time window. The frequency domain feature pairs and time domain feature values ​​of the same time window are combined to form the single-window acoustic feature vector of that time window. The single-window acoustic feature vectors of all time windows in the corresponding acoustic signal history segment of the spatial region are arranged in chronological order of the time windows to obtain the acoustic feature parameter combination.

6. The production management method for supercritical foamed materials used for the protection of new energy batteries according to claim 5, characterized in that, The specific process of signal separation and processing is as follows: When the foaming vessel is in an unloaded state without the injection of supercritical fluid and without the placement of polymer matrix, an unloaded depressurization operation with the same depressurization rate and the same initial depressurization pressure as a normal production batch is performed. Multiple acoustic wave sensors deployed on the inner wall of the foaming vessel collect acoustic wave signals throughout the entire unloaded depressurization operation process. The collected acoustic wave signals are used as the unloaded depressurization background acoustic wave signals corresponding to each acoustic wave sensor and stored. For any spatial region in a batch, the background acoustic signal of the acoustic sensor corresponding to the spatial region is obtained under no-load pressure relief. The acoustic amplitude of each sampling time in the original acoustic signal history segment is subtracted point by point from the acoustic amplitude of the background acoustic signal of the no-load pressure relief at the same sampling time to obtain the difference acoustic signal segment of the spatial region in the batch. The difference acoustic signal segment is input into a bandpass filter. The lower cutoff frequency of the bandpass filter is set to the lowest dominant frequency of the acoustic signal generated during the bubble merging process, and the upper cutoff frequency of the bandpass filter is set to the highest dominant frequency of the acoustic signal generated during the bubble nucleation process. The bandpass filter filters out signal components below the lower cutoff frequency and signal components above the upper cutoff frequency, retaining signal components between the lower and upper cutoff frequencies. The filtered difference acoustic signal segment is then used as the valid acoustic signal segment for that spatial region in that batch.

7. The production management method for supercritical foamed materials used for the protection of new energy batteries according to claim 5, characterized in that, In S3, the specific extraction process for the degree of deviation is as follows: The acoustic feature parameter combination sequence of each spatial region within a high-uniformity batch is subjected to intra-batch normalization transformation. The single-window acoustic feature vector at the same time window position in the acoustic feature parameter combination sequence of each spatial region is subtracted from the mean of the single-window acoustic feature vectors of all spatial regions in the batch at the same time window position, and then divided by the standard deviation of the single-window acoustic feature vectors of all spatial regions in the batch at the same time window position to obtain the normalized single-window acoustic feature vector of each spatial region at each time window position. For any two spatial regions within a highly uniform batch, the difference between the standardized single-window acoustic feature vectors of the two spatial regions at the same time window position is squared to obtain a square value. The square values ​​at all time window positions are summed, and the square root of the sum is taken as the degree of deviation between the two spatial regions.

8. The production management method for supercritical foamed materials used for the protection of new energy batteries according to claim 1, characterized in that, S5 further includes generating a batch sampling decision instruction to be executed according to a preset conventional sampling plan if the deviation between any two spatial regions does not exceed the reference interval library of acoustic feature deviations between regions.

Images