Artificial intelligence-based wafer throwing rubber floor particle mixing uniformity regulation method

Abstract

This invention relates to the field of data processing, specifically to an artificial intelligence-based method for controlling the mixing uniformity of plywood flooring particles. The method includes: acquiring temperature and vibration data during the mixing process of plywood flooring particles; determining the period of the vibration data, dividing the vibration data into period segments, and calculating vibration characteristic values ​​based on the vibration data in each period segment; determining a temperature change rate sequence based on reference temperature data, and calculating temperature characteristic values ​​based on the temperature change rate sequence and at least two comparative temperature data; acquiring spatiotemporal correlation characteristic values ​​of the temperature and vibration data; acquiring feature vectors; training a random forest model; inputting the vector to be analyzed into the trained random forest model, outputting data labels through the trained random forest model, and adjusting the mixer speed based on the data labels. This invention can improve the mixing uniformity of rubber flooring during substrate mixing.

Description

Technical Field

[0001] This invention relates to the field of data processing, and more specifically to a method for controlling the mixing uniformity of granules in polished rubber flooring based on artificial intelligence. Background Technology

[0002] In the manufacturing of polished rubber flooring, there are currently two main process routes: one is the traditional "surface scattering-hot pressing" method, which only achieves a shallow physical embedding of anti-slip particles, which easily leads to problems such as particle detachment, numerous surface burrs, dirt accumulation in interface gaps, and wear and failure of the decorative layer, and cannot guarantee the permanent decorative effect of the product; the other is the improved "through-body scattering" method, which premixes particles inside the substrate and adds a scattering layer to the surface to extend the visual effect, but still fails to solve the fundamental problems of surface particle burrs and weak bonding force, and the uniformity of internal particle distribution cannot be accurately monitored and controlled.

[0003] Compared to the first manufacturing process, the second process mentioned above has made fundamental improvements in the premixing of the substrate. However, the problem of uneven substrate mixing still exists. Although it improves visual continuity to some extent, it fails to fundamentally solve the problem of weak surface particle adhesion. Furthermore, due to the lack of effective online monitoring methods, it is difficult to accurately control the uniformity of particle distribution within the substrate, resulting in unstable product quality and low yield. Summary of the Invention

[0004] This invention provides an artificial intelligence-based method for controlling the mixing uniformity of granules in polished rubber flooring, in order to solve existing problems.

[0005] The artificial intelligence-based method for controlling the mixing uniformity of granulated rubber flooring particles of the present invention adopts the following technical solution:

[0006] One embodiment of the present invention provides a method for controlling the mixing uniformity of plywood flooring particles based on artificial intelligence, the method comprising the following steps:

[0007] Acquire temperature and vibration data during the mixing process of the granulated rubber flooring particles, wherein the temperature data includes baseline temperature data and at least two comparative temperature data;

[0008] Determine the period of the vibration data, divide the vibration data according to the period to obtain period segments, and calculate the vibration characteristic value of the vibration data based on the vibration data in each period segment;

[0009] Determine the temperature change rate sequence based on the reference temperature data, and calculate the temperature characteristic value of the temperature data based on the temperature change rate sequence and at least two comparative temperature data.

[0010] Acquire the spatiotemporal correlation feature values ​​of temperature and vibration data;

[0011] Acquire historical temperature data and historical vibration data, calculate historical vibration data variance, historical vibration eigenvalues, historical temperature data variance, historical temperature eigenvalues, and historical spatiotemporal correlation eigenvalues ​​based on the historical temperature data and historical vibration data, and form an eigenvector;

[0012] The random forest model is trained by using the feature vector as input and the corresponding data label as output. The data label is the binary label of the feature vector.

[0013] The variance of vibration data, vibration eigenvalues, variance of temperature data, temperature eigenvalues, and spatiotemporal correlation eigenvalues ​​are used to form an analysis vector. This analysis vector is then input into a trained random forest model, which outputs data labels. These data labels are used to adjust the mixer's rotation speed.

[0014] Optionally, the period of the vibration data is determined, specifically including:

[0015] Perform a Fourier transform on the vibration data to obtain the Fourier transform result;

[0016] The maximum amplitude is obtained from the Fourier transform result, and the reciprocal of the maximum amplitude is used to determine the period.

[0017] Optionally, based on the vibration data in each period segment, the vibration characteristic values ​​of the vibration data are calculated, specifically including:

[0018] The vibration data in each period segment are clustered separately to obtain the vibration data category for each period segment;

[0019] For each period segment, the maximum vibration value in the vibration data category with the largest amount of data is determined as the basic number of the period segment;

[0020] The difference between each vibration value and the basic number in the period segment is obtained, and the ratio of the difference to the basic number is used to determine the vibration anomaly of each vibration value.

[0021] Vibration values ​​within a periodic segment that exhibit abnormal vibration exceeding a preset abnormal vibration threshold are defined as abnormal vibration values.

[0022] For the entire periodic segment, the time point of each abnormal vibration value is determined as the abnormal time point;

[0023] Obtain the ratio of the frequency of each abnormal time point to the number of periodic segments, and determine the maximum ratio as the vibration characteristic value.

[0024] Optionally, the time point for each abnormal vibration value is obtained as follows:

[0025] The time point of the nth abnormal vibration value in the mth period segment is determined by the difference between the time corresponding to the nth abnormal vibration value in the mth period segment and the time corresponding to the first vibration value in the mth period segment.

[0026] Obtain the time point for each abnormal vibration value in the m-th period segment;

[0027] Obtain the time point for each abnormal vibration value in each period segment.

[0028] Optionally, a temperature change rate sequence is determined based on reference temperature data, and temperature characteristic values ​​of the temperature data are calculated based on the temperature change rate sequence and at least two comparative temperature data, specifically including:

[0029] The absolute value of the difference between the a-th temperature value and the (a-1)-th temperature value in the reference temperature data is taken, and the ratio of the absolute value to the (a-1)-th temperature value is determined as the temperature change rate of the a-th temperature value. The a-th temperature value is not the first temperature value in the reference temperature data.

[0030] The sequence of temperature change rates for each temperature value in the reference temperature data is defined as the temperature change rate sequence.

[0031] The maximum value in the temperature change rate sequence is determined as the reference temperature change rate;

[0032] Obtain the temperature change rate sequence for each comparative temperature data point;

[0033] In each comparison temperature change rate sequence, elements that are greater than the baseline temperature change rate are identified as outlier values.

[0034] The temperature value corresponding to the abnormal element value is identified as the abnormal temperature value;

[0035] For each abnormal temperature value, calculate the difference between the comparative temperature change rate and the baseline temperature change rate, and determine the temperature abnormality of each abnormal temperature value by the ratio of the difference to the baseline temperature change rate.

[0036] The average of all abnormal temperature values ​​is used to obtain the temperature characteristic value.

[0037] Optionally, the spatiotemporal correlation feature values ​​of temperature data and vibration data are obtained, specifically including:

[0038] The mean value of temperature values ​​in the baseline temperature data is determined as the normal temperature value, and the ratio of abnormal temperature values ​​to normal temperature values ​​is determined as the temperature anomaly rate.

[0039] In the comparison temperature data, the abnormal temperature value in the first comparison temperature data is determined as the first abnormal temperature data, and the abnormal temperature value in the second comparison temperature data is determined as the second abnormal temperature data.

[0040] The temperature anomaly rate of each first abnormal temperature data is determined as the first temperature anomaly rate, and the temperature anomaly rate of each second abnormal temperature data is determined as the second temperature anomaly rate.

[0041] The first temperature anomaly rate and the second temperature anomaly rate are respectively determined as the matching set of the extended KM matching algorithm, and the ratio of the temperature anomaly rates is determined as the weight of the edge of the extended KM matching algorithm. The extended KM matching algorithm is run to obtain the matching relationship.

[0042] The first temperature anomaly rate and the second temperature anomaly rate are clustered separately to obtain the first cluster set and the second cluster set;

[0043] Based on the matching relationship, determine the matching relationship between the first cluster set and the second cluster set, and based on the matching relationship between the first cluster set and the second cluster set, determine the matching cluster pairs, and determine the set of matching cluster pairs as the matching set;

[0044] For each matching cluster pair in the matching set, obtain the first and second comparative temperature data of the matching cluster pair;

[0045] The time point in the comparative temperature data for each target temperature anomaly rate is determined as the target anomaly time point, where the target temperature anomaly rate is the first temperature anomaly rate of the first comparative temperature data and the second temperature anomaly rate of the second comparative temperature data.

[0046] Calculate the spatiotemporal correlation feature value based on the target anomaly time point.

[0047] Optionally, based on the matching relationship, the matching relationship between the first cluster set and the second cluster set is determined, and based on the matching relationship between the first cluster set and the second cluster set, matching cluster pairs are determined. The set composed of matching cluster pairs is determined as the matching set, specifically including:

[0048] Based on the matching relationship, the second temperature anomaly rate corresponding to the cth first temperature anomaly rate in the bth cluster of the first cluster set is determined as the matching element, and the clusters of the matching element in the second cluster set are determined as candidate matching clusters.

[0049] Obtain the candidate matching clusters for the b-th cluster in the first set of clusters;

[0050] The ratio of the number of matching elements in the candidate matching cluster to the number of elements in the candidate matching cluster is determined as the matching value of the candidate matching cluster.

[0051] The candidate matching clusters whose matching value is greater than the preset matching threshold are determined as the matching clusters of the b-th cluster in the first cluster set, and the b-th cluster in the first cluster set and the matching cluster are determined as a matching cluster pair;

[0052] Any data point in the comparative temperature data is designated as the first comparative temperature data point, and any data point other than the first comparative temperature data point is designated as the second comparative temperature data point. Matching cluster pairs are obtained, and the set of matching cluster pairs is determined as the matching set.

[0053] Optionally, spatiotemporal correlation feature values ​​are calculated based on the target anomaly time point, specifically including:

[0054] The time points in the vibration data corresponding to the abnormal time points where the vibration characteristic value is greater than the preset vibration characteristic threshold are determined as abnormal vibration time points;

[0055] Obtain the target time point of the anomaly;

[0056] Calculate the absolute value of the difference between the z-th target anomaly time point and each vibration anomaly time point to obtain the matching duration set of the z-th target anomaly time point;

[0057] The vibration anomaly time point corresponding to the element in the matching time set that is less than the preset time threshold is determined as the matching time point of the z-th target anomaly time point;

[0058] Obtain the matching time point for each target anomaly time point;

[0059] Obtain the ratio of the number of matching time points to the number of vibration anomaly time points, and determine the maximum ratio as the spatiotemporal correlation feature value.

[0060] Optionally, the first temperature anomaly rate and the second temperature anomaly rate are obtained by sorting the temperature anomaly rates in ascending order. When the first temperature anomaly rate is greater than the second temperature anomaly rate, the temperature anomaly rate is the ratio of the second temperature anomaly rate to the first temperature anomaly rate. When the first temperature anomaly rate is less than or equal to the second temperature anomaly rate, the temperature anomaly rate is the ratio of the first temperature anomaly rate to the second temperature anomaly rate.

[0061] This invention proposes an artificial intelligence-based system for controlling the mixing uniformity of granulated rubber flooring particles, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the artificial intelligence-based method for controlling the mixing uniformity of granulated rubber flooring particles.

[0062] The beneficial effects of the technical solution of the present invention are:

[0063] In this embodiment of the invention, by using multimodal sensor data fusion and spatiotemporal correlation analysis, early and accurate identification and active closed-loop control of particle agglomeration during the mixing process are achieved, thereby significantly improving the consistency of anti-slip performance and quality stability of the finished rubber flooring, while reducing the scrap rate and production costs. Attached Figure Description

[0064] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0065] Figure 1 A flowchart illustrating an artificial intelligence-based method for controlling the mixing uniformity of granulated rubber flooring particles, as provided in an embodiment of the present invention.

[0066] Figure 2 This is a structural diagram of an artificial intelligence-based particle mixing uniformity control system for polished rubber flooring provided in one embodiment of the present invention. Detailed Implementation

[0067] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the artificial intelligence-based method for controlling the mixing uniformity of plywood flooring particles proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0068] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0069] The following describes in detail, with reference to the accompanying drawings, the specific scheme of the artificial intelligence-based method for controlling the mixing uniformity of granules in sheet rubber flooring provided by the present invention.

[0070] This invention provides a method and system for controlling the mixing uniformity of granulated rubber flooring particles based on artificial intelligence. Please refer to [link to relevant documentation]. Figure 1 The diagram illustrates a flowchart of an artificial intelligence-based method for controlling the mixing uniformity of plywood flooring particles according to an embodiment of the present invention. The method includes the following steps:

[0071] S101. Obtain temperature and vibration data during the mixing process of the granulated rubber flooring particles, wherein the temperature data includes reference temperature data and at least two comparison temperature data.

[0072] For example, before manufacturing polished rubber flooring, it is necessary to prepare the substrate required for manufacturing. The preparation process is as follows:

[0073] Weighing and refining: According to the preset formula, use a high-precision scale to weigh the rubber base material and the anti-slip granular rubber material separately, and then refine them.

[0074] Anti-slip granule processing: The anti-slip granule rubber compound is colored according to decorative requirements, followed by extrusion and pre-vulcanization. The pre-vulcanized rubber compound is cut into strips, then crushed and screened to ensure that its particle size and shape meet the design requirements.

[0075] Substrate preparation: Slice the refined rubber substrate and weigh it accurately for later use.

[0076] The pretreated rubber base material, anti-slip granules, color masterbatch, and all other solid raw materials are fed into an open mill for physical mixing. The purpose of this stage is to achieve a preliminary uniform distribution of the components on a macroscopic scale.

[0077] To monitor mixing uniformity, this embodiment deploys sensors based on the following physical principles:

[0078] Since the density of anti-slip particles is usually higher than that of the base material particles, if the high-density, small-particle-size anti-slip particles form agglomerates during the mixing process, these agglomerates will mainly settle under the action of gravity and centrifugal force, and periodically slide and roll along the bottom area of ​​the open mill and the junction of the bottom and sidewall of the pot. This motion characteristic will simultaneously cause local temperature anomalies and periodic mechanical impacts.

[0079] To capture the above signals, this embodiment installs the following sensor system on the open mill (taking an open mill or any mixing device as an example):

[0080] (1) Motion and vibration monitoring:

[0081] Rotary encoder: Installed on the main shaft of the open mill roller, it is used to obtain the precise angular position of the roller in real time, providing a reference for the period division of the vibration signal (if it is a mixing device, it is used to obtain the angular position of the mixing blade).

[0082] High-frequency vibration acceleration sensor: installed on the bearing housing of the open mill, used to collect high-frequency impact vibration signals generated by the periodic impact of particle agglomerates.

[0083] (2) Temperature monitoring network (RTD deployment):

[0084] The monitoring focus is on the bottom and lower sidewall areas where particle aggregation is prone to occur, while a reference point is set to obtain the background temperature.

[0085] Location 1 (center point of the base plate):

[0086] Installation location: The exact center of the bottom plate of the open mill.

[0087] Installation method: Install a sturdy temperature measuring sleeve vertically upward from the center of the base plate, with the top of the sleeve flush with the inner wall of the vessel bottom, and insert an armored RTD inside.

[0088] Function: Directly monitors the temperature of the central area at the bottom, which is a key point for capturing bottom aggregation.

[0089] Location 2 (path point at the bottom of the side wall):

[0090] Installation location: At least 3 RTDs are evenly spaced along the sweeping circumference of the roller at the bottom of the side wall of the open mill (the more placement points, the more accurate the positioning).

[0091] Function: To monitor the abnormal temperature trajectory generated when particle clusters move periodically along the bottom of the sidewall.

[0092] Location 3 (Reference point in the middle of the side wall):

[0093] Installation location: Install an RTD at the midpoint of the side wall.

[0094] Function: Provides a normal background temperature value as a benchmark for comparison with the temperature data at location one and location two, thereby more accurately identifying abnormal temperature values ​​caused by particle agglomeration.

[0095] The real-time data collected by the aforementioned sensor network is defined as follows:

[0096] Vibration data: acquired by a high-frequency vibration acceleration sensor.

[0097] Reference temperature data: acquired from the RTD at location 3, used to characterize the normal temperature background of the system.

[0098] Temperature data for comparison: acquired from RTD arrays at positions one and two, serving as the primary data source for anomaly analysis and spatiotemporal correlation feature calculation.

[0099] S102. Determine the period of the vibration data, divide the vibration data according to the period to obtain period segments, and calculate the vibration characteristic value of the vibration data based on the vibration data in each period segment.

[0100] In this embodiment, determining the period of vibration data specifically includes:

[0101] Perform a Fourier transform on the vibration data to obtain the Fourier transform result;

[0102] The maximum amplitude is obtained from the Fourier transform result, and the reciprocal of the maximum amplitude is used to determine the period.

[0103] Based on the vibration data in each period segment, the vibration characteristic values ​​of the vibration data are calculated, specifically including:

[0104] The vibration data in each period segment are clustered separately to obtain the vibration data category for each period segment;

[0105] For each period segment, the maximum vibration value in the vibration data category with the largest amount of data is determined as the basic number of the period segment;

[0106] The difference between each vibration value and the basic number in the period segment is obtained, and the ratio of the difference to the basic number is used to determine the vibration anomaly of each vibration value.

[0107] Vibration values ​​within a periodic segment that exhibit abnormal vibration exceeding a preset abnormal vibration threshold are defined as abnormal vibration values.

[0108] For the entire periodic segment, the time point of each abnormal vibration value is determined as the abnormal time point;

[0109] Obtain the ratio of the frequency of each abnormal time point to the number of periodic segments, and determine the maximum ratio as the vibration characteristic value.

[0110] The time point for each abnormal vibration value is obtained as follows:

[0111] The time point of the nth abnormal vibration value in the mth period segment is determined by the difference between the time corresponding to the nth abnormal vibration value in the mth period segment and the time corresponding to the first vibration value in the mth period segment.

[0112] Obtain the time point for each abnormal vibration value in the m-th period segment;

[0113] Obtain the time point for each abnormal vibration value in each period segment.

[0114] For example, vibration and temperature data during the mixing process can be used to indicate mixing uniformity. However, vibration signals reflect the action of "force," not the "distribution" itself. It is difficult to distinguish whether abnormal vibration is caused by particle agglomeration or by mechanical failures of the equipment itself (such as bearing wear, loose impellers, or dynamic imbalance). Similarly, uneven temperature distribution does not necessarily stem from particle agglomeration; it may also be caused by uneven heaters, improper cooling coil positioning, or uneven feed temperature.

[0115] Therefore, in constructing the feature values ​​of the classifier for the random forest model, this embodiment not only extracts the vibration feature values ​​and temperature feature values ​​that reflect the uniformity of mixing, but also further introduces the spatiotemporal correlation feature values ​​of the two to improve the recognition performance and robustness of the classifier.

[0116] The principle is that vibration signals can accurately indicate the "moment" and "intensity" of abnormal dynamic events, while temperature data can clearly present the "location" and "range" of abnormal physical property distribution. When both detect anomalies simultaneously and show a significant correlation in time and space—for example, the abnormally high temperature area in the temperature image coincides with the trajectory of the stirring paddle, and a characteristic impact pulse appears synchronously in the vibration signal at the instant the area approaches the stirring paddle—a strong chain of evidence can be formed, thus determining with high confidence that the anomaly is a real particle agglomeration.

[0117] It should be noted that in an open mill, because the density of the anti-slip particles is greater than that of the base material particles, they often settle in the lower part of the mixing chamber during the mixing process. Thermal infrared sensors typically only monitor the upper surface of the equipment, as the sides are blocked by the outer casing, thus they cannot directly detect temperature anomalies in the lower internal areas.

[0118] During the mixing process in an open mill, the vibration signals generated by the equipment drive system typically exhibit a fundamental frequency and its harmonic components that are strictly synchronized with the rotor (or roller) rotation cycle. This is considered normal hydrodynamic and mechanical vibration background. When there are insufficiently dispersed agglomerates with a certain structural strength in the mixed material, these agglomerates will periodically compress or impact the rotating rotor (or roller). In the time domain, this interaction manifests as a transient impact pulse with a significantly high amplitude superimposed at a specific phase of each rotation cycle on a stable periodic background vibration waveform.

[0119] Therefore, the vibration data is subjected to Fourier transform to transform it into the frequency domain space with frequency on the horizontal axis and amplitude on the vertical axis. The reciprocal of the frequency corresponding to the maximum amplitude is taken as the period. The vibration data is divided into multiple period segments by this period.

[0120] For the vibration data in each cycle segment, the DBSCAN density clustering method is used to cluster the vibration data in each cycle segment, obtaining multiple vibration data categories. The maximum vibration value in the category with the largest amount of data is taken as the base value for that cycle segment. For each other vibration data in that cycle segment, the absolute value of the difference between it and the base value is calculated, and the ratio of this absolute value to the base value is determined as the vibration anomaly of that vibration data. Vibration values ​​with an anomaly value greater than a preset vibration anomaly threshold (e.g., 0.7) are determined as abnormal vibration values.

[0121] The key point is that if a certain time location is identified as an abnormal vibration value in vibration data of different cycles, that is, a high-amplitude impact occurs at the same time in each cycle, this phenomenon may be caused by persistent anti-slip particle clumps.

[0122] For the vibration data of each period segment, obtain the time point corresponding to each abnormal vibration value. Here, the time point refers to the time difference from the start time of the period segment to the time when the abnormal vibration value appears. Thus, all abnormal time points for all period segments can be obtained.

[0123] In all periodic segments, the frequency of occurrence at each time point is counted, and the ratio of this frequency to the total number of periodic segments is calculated. The largest ratio is identified and denoted as b1. The larger the b1 value, the stronger the periodic recurrence of the impact event occurring at that time point, and therefore the more likely it is caused by periodically passing anti-slip particle clusters (rather than random noise or transient interference). Finally, this b1 value is used as the vibration characteristic value.

[0124] Optionally, the preset vibration anomaly threshold is 0.7 in a preferred embodiment. The value of the preset vibration anomaly threshold can be modified according to actual needs and historical experience, and no specific numerical limit is imposed here.

[0125] S103. Determine the temperature change rate sequence based on the reference temperature data, and calculate the temperature characteristic value of the temperature data based on the temperature change rate sequence and at least two comparative temperature data.

[0126] In this embodiment, a temperature change rate sequence is determined based on reference temperature data, and temperature characteristic values ​​of the temperature data are calculated based on the temperature change rate sequence and at least two comparative temperature data, specifically including:

[0127] The absolute value of the difference between the a-th temperature value and the (a-1)-th temperature value in the reference temperature data is taken, and the ratio of the absolute value to the (a-1)-th temperature value is determined as the temperature change rate of the a-th temperature value. The a-th temperature value is not the first temperature value in the reference temperature data.

[0128] The sequence of temperature change rates for each temperature value in the reference temperature data is defined as the temperature change rate sequence.

[0129] The maximum value in the temperature change rate sequence is determined as the reference temperature change rate;

[0130] Obtain the temperature change rate sequence for each comparative temperature data point;

[0131] In each comparison temperature change rate sequence, elements that are greater than the baseline temperature change rate are identified as outlier values.

[0132] The temperature value corresponding to the abnormal element value is identified as the abnormal temperature value;

[0133] For each abnormal temperature value, calculate the difference between the comparative temperature change rate and the baseline temperature change rate, and determine the temperature abnormality of each abnormal temperature value by the ratio of the difference to the baseline temperature change rate.

[0134] The average of all abnormal temperature values ​​is used to obtain the temperature characteristic value.

[0135] For example, in an open mill, the high-speed friction between anti-slip particles generates heat, which can easily lead to clumping due to surface stickiness. In this case, abnormal changes in the temperature values ​​of RTDs (resistance temperature sensors) at different locations can be used to identify potential clumping of anti-slip particles.

[0136] Temperature data at location 3, located within the typical normal range, is used as the baseline temperature data. For this sequence, the absolute value of the difference between each subsequent temperature value and the preceding temperature value is calculated, and the ratio of this absolute value to the preceding temperature value is determined as the temperature change rate at that moment. This yields a temperature change rate sequence, used to characterize the normal temperature change pattern in the open mill. The maximum value in this temperature change rate sequence is recorded as the baseline temperature change rate.

[0137] For the temperature data (i.e., comparative temperature data) obtained from locations 1 and 2 that need to be monitored and analyzed, the same method is used to calculate their respective comparative temperature change rate sequences. Element values ​​in these two sequences that exceed the baseline temperature change rate are identified as anomalous element values. The temperature values ​​corresponding to these anomalous element values ​​in the original temperature sequences are recorded as anomalous temperature values.

[0138] For each anomalous temperature value, the difference between its corresponding rate of temperature change and the reference rate of temperature change is calculated. This difference is then divided by the reference rate of temperature change to obtain the temperature anomaly of that anomalous temperature value. The larger this ratio, the more significant the temperature anomaly at that location. Finally, the average of the temperature anomalies of all anomalous temperature values ​​is calculated and denoted as b2. This b2 value is then used as the final temperature characteristic value.

[0139] S104. Obtain the spatiotemporal correlation feature values ​​of temperature data and vibration data.

[0140] In this embodiment, obtaining the spatiotemporal correlation feature values ​​of temperature data and vibration data specifically includes:

[0141] The mean value of temperature values ​​in the baseline temperature data is determined as the normal temperature value, and the ratio of abnormal temperature values ​​to normal temperature values ​​is determined as the temperature anomaly rate.

[0142] In the comparison temperature data, the abnormal temperature value in the first comparison temperature data is determined as the first abnormal temperature data, and the abnormal temperature value in the second comparison temperature data is determined as the second abnormal temperature data.

[0143] The temperature anomaly rate of each first abnormal temperature data is determined as the first temperature anomaly rate, and the temperature anomaly rate of each second abnormal temperature data is determined as the second temperature anomaly rate.

[0144] The first temperature anomaly rate and the second temperature anomaly rate are respectively determined as the matching set of the extended KM matching algorithm, and the ratio of the temperature anomaly rates is determined as the weight of the edge of the extended KM matching algorithm. The extended KM matching algorithm is run to obtain the matching relationship.

[0145] The first temperature anomaly rate and the second temperature anomaly rate are clustered separately to obtain the first cluster set and the second cluster set;

[0146] Based on the matching relationship, determine the matching relationship between the first cluster set and the second cluster set, and based on the matching relationship between the first cluster set and the second cluster set, determine the matching cluster pairs, and determine the set of matching cluster pairs as the matching set;

[0147] For each matching cluster pair in the matching set, obtain the first and second comparative temperature data of the matching cluster pair;

[0148] The time point in the comparative temperature data for each target temperature anomaly rate is determined as the target anomaly time point, where the target temperature anomaly rate is the first temperature anomaly rate of the first comparative temperature data and the second temperature anomaly rate of the second comparative temperature data.

[0149] Based on the target abnormal time point, the spatiotemporal correlation feature value is calculated. The first temperature abnormality rate and the second temperature abnormality rate are obtained by sorting the temperature abnormality rates in ascending order. When the first temperature abnormality rate is greater than the second temperature abnormality rate, the temperature abnormality rate is the ratio of the second temperature abnormality rate to the first temperature abnormality rate. When the first temperature abnormality rate is less than or equal to the second temperature abnormality rate, the temperature abnormality rate is the ratio of the first temperature abnormality rate to the second temperature abnormality rate.

[0150] Based on the matching relationship, determine the matching relationship between the first cluster set and the second cluster set, and based on the matching relationship between the first cluster set and the second cluster set, determine the matching cluster pairs. The set of matching cluster pairs is determined as the matching set, specifically including:

[0151] Based on the matching relationship, the second temperature anomaly rate corresponding to the cth first temperature anomaly rate in the bth cluster of the first cluster set is determined as the matching element, and the clusters of the matching element in the second cluster set are determined as candidate matching clusters.

[0152] Obtain the candidate matching clusters for the b-th cluster in the first set of clusters;

[0153] The ratio of the number of matching elements in the candidate matching cluster to the number of elements in the candidate matching cluster is determined as the matching value of the candidate matching cluster.

[0154] The candidate matching clusters whose matching value is greater than the preset matching threshold are determined as the matching clusters of the b-th cluster in the first cluster set, and the b-th cluster in the first cluster set and the matching cluster are determined as a matching cluster pair;

[0155] Any data point in the comparative temperature data is designated as the first comparative temperature data point, and any data point other than the first comparative temperature data point is designated as the second comparative temperature data point. Matching cluster pairs are obtained, and the set of matching cluster pairs is determined as the matching set.

[0156] Based on the target anomaly time point, calculate the spatiotemporal correlation feature value, specifically including:

[0157] The time points in the vibration data corresponding to the abnormal time points where the vibration characteristic value is greater than the preset vibration characteristic threshold are determined as abnormal vibration time points;

[0158] Obtain the target time point of the anomaly;

[0159] Calculate the absolute value of the difference between the z-th target anomaly time point and each vibration anomaly time point to obtain the matching duration set of the z-th target anomaly time point;

[0160] The vibration anomaly time point corresponding to the element in the matching time set that is less than the preset time threshold is determined as the matching time point of the z-th target anomaly time point;

[0161] Obtain the matching time point for each target anomaly time point;

[0162] Obtain the ratio of the number of matching time points to the number of vibration anomaly time points, and determine the maximum ratio as the spatiotemporal correlation feature value.

[0163] For example, if the occurrence of abnormal temperatures is periodic, such as caused by the periodic sliding and rolling of anti-slip particle clusters along the junction of the bottom and sidewall of the reactor, it can be identified by analyzing the periodic movement of the temperature anomalies at locations 1 and 2 over time.

[0164] Vibration signals reflect the action of "force," making it difficult to distinguish whether the source is particle agglomeration or mechanical failure of the equipment itself (such as bearing wear or dynamic imbalance). Similarly, uneven temperature does not necessarily originate from particle agglomeration; it may also be caused by uneven heaters, improper cooling coil positioning, or uneven feed temperature.

[0165] When both vibration and temperature data indicate anomalies, and these anomalies show a significant correlation in time and space, a high degree of confidence in the occurrence of particle agglomeration can be established. Specifically, when a thermal anomaly area is detected, if the vibration data synchronously shows characteristic and relatively obvious abnormal impacts at the instant the area approaches the agitator (or roller), then the thermal anomaly area can be identified as a genuine particle agglomeration that is periodically interacting with the equipment.

[0166] For all temperature values ​​at location 3, calculate their arithmetic mean and record this mean as the normal temperature value. For each RTD to be analyzed (specifically, the RTDs at locations 1 and 2), obtain all its abnormal temperature values. For each abnormal temperature value, calculate the difference between it and the normal temperature value, and divide this difference by the normal temperature value. The resulting ratio is recorded as the temperature anomaly rate for that abnormal temperature value.

[0167] KM matching (Kuhn-Munkres algorithm, also known as the Hungarian algorithm) is a classic algorithm for finding the maximum weight perfect matching in a bipartite graph. Its core is to find a one-to-one matching relationship between each value on the left and each value on the right.

[0168] The extended KM matching algorithm is an optimized version of the KM matching algorithm. The difference between the extended and KM matching algorithms is that the KM matching algorithm is mainly applied to matching two subsets with exactly the same number of elements, while the extended KM matching algorithm can be applied to matching two subsets with different numbers of elements.

[0169] Specifically, the extended KM matching algorithm has been well-established in practice. This section mainly describes the main differences between the extended KM matching algorithm and the KM matching algorithm, without providing a detailed explanation of the extended KM matching algorithm. For details, please refer to the practical applications of how the KM matching algorithm matches different numbers of subsets.

[0170] Specifically, in this embodiment, the application is as follows: For any two RTDs to be compared, denoted as A and B, firstly, all temperature anomaly rates in the data of both are obtained. The set of all temperature anomaly rates (first temperature anomaly rate) in RTD A (the temperature data collected by RTD A is denoted as the first comparison temperature data) is used as the left node of the bipartite graph, and the set of all temperature anomaly rates (second temperature anomaly rate) in RTD B (the temperature data collected by RTD B is denoted as the second comparison temperature data) is used as the right node. Full connections are constructed between the left and right nodes, meaning that each node on the left is connected to each node on the right by an edge. The weight of each edge is defined as the ratio of the smaller to the larger of the two temperature anomaly rates connected by the edge (this ratio ranges from 0 to 1; a larger value indicates a more similarity between the two anomaly rates). Subsequently, the KM matching algorithm is run, following the maximum matching principle (i.e., seeking the maximum sum of the weights of all matching edges), thereby obtaining a one-to-one matching relationship between the left and right nodes. The final output is a set of multiple matching pairs (i.e., matching relationships), where each matching pair contains a temperature anomaly rate element from RTD A and a temperature anomaly rate element from RTD B.

[0171] Furthermore, the data was preprocessed and nodes were defined before constructing the matching graph:

[0172] Sorting rules: All temperature anomaly rates in RTD A are sorted in ascending order of their numerical values; all temperature anomaly rates in RTDB are also sorted using the same method. This sorting facilitates the orderly processing of subsequent algorithms, but the matching relationship is based on the numerical values ​​themselves and does not depend on their sorted positions.

[0173] Node generation rules: For each RTD, a corresponding node is generated for each moment in its measurement time series where a temperature anomaly rate exists. Importantly, node generation is based on every anomalous moment in the original time series. Therefore, if the same RTD exhibits the same temperature anomaly rate value at any number of different moments, each such moment will generate an independent node, without deduplication or retention based solely on identical values. This ensures the integrity of the temporal dimension information, enabling subsequent matching to reflect the temporal distribution and correspondence of anomalous events.

[0174] Cluster all temperature anomaly rates on the left (corresponding to RTD A) to obtain multiple first-class clusters (denoted as cluster 1); cluster all temperature anomaly rates on the right (corresponding to RTD B) to obtain multiple second-class clusters (denoted as cluster 2).

[0175] Based on this, if the matching points of most elements in a cluster 1 (i.e., the temperature anomaly rate from RTD B matched by the extended KM matching algorithm) are concentrated in a specific cluster 2, then this pattern is considered to correspond to the periodic movement process of the same anti-slip particle cluster from position A to position B.

[0176] For each cluster 1, obtain the matching pairs obtained by KM matching for each temperature anomaly rate element in the cluster, and then extract the matching elements located on the RTD B side from these matching pairs. Count the clusters 2 to which all these matching elements belong, and record these clusters as candidate matching clusters 2 of cluster 1.

[0177] For each candidate matching cluster 2, calculate the proportion of matching elements in that cluster to the total number of elements in that cluster, and denot this as the matching value of the candidate matching cluster. If the matching value of a candidate matching cluster 2 is greater than a preset matching threshold (e.g., 0.7), then this high proportion of matching relationships is considered to likely correspond to the back-and-forth movement of an anti-slip particle cluster between RTD A and RTD B. The matching values ​​that satisfy this condition and their corresponding clusters 1 and 2 are retained, together forming a matching cluster pair.

[0178] Obtain the RTDs (i.e., RTD A and RTD B) associated with all retained matching cluster pairs. For any RTD, extract the time points corresponding to all temperature anomaly rates in its matching cluster pair, and then summarize them to obtain the set of anomaly time points for all related RTDs.

[0179] Based on physical mechanisms: if these temperature anomalies are indeed caused by a moving anti-slip particle cluster, then when the roller (or agitator) moves close to the RTD detection position (i.e., near the anomaly time point), it is expected to collide or compress with the anti-slip particle cluster, thereby triggering a large vibration value (i.e., an abnormal vibration value) in the vibration data. Therefore, the presence of anti-slip particle clusters can be aided in determining whether the large vibration value always occurs near the temperature anomaly time point.

[0180] All time points where the vibration characteristic value b1 is greater than a preset vibration characteristic threshold (e.g., 0.7) are identified and defined as vibration anomaly time points. For each vibration anomaly time point, the absolute value of the difference between it and each obtained temperature anomaly time point is calculated. Anomaly time points whose absolute difference is less than a preset duration threshold (e.g., 1 second) are marked as their corresponding time points. The total number of corresponding time points found for all vibration anomaly time points is counted, and the ratio of this total number to the total number of vibration anomaly time points is calculated, denoted as g. The larger the ratio g, the higher the temporal synergy between vibration anomalies and temperature anomalies, and therefore the greater the probability of the presence of anti-slip particle clusters.

[0181] The calculated maximum ratio g (or the representative ratio g selected from multiple matching sets) is determined as the final spatiotemporal correlation feature value. This feature value is used to quantitatively characterize the degree of consistency between temperature features and vibration features in their occurrence sequence.

[0182] Optionally, in a preferred embodiment, the preset matching threshold and the preset vibration characteristic threshold can be set to 0.7 and 0.7, respectively. The preset matching threshold and the preset vibration characteristic threshold can be modified according to actual needs and historical experience, and no specific numerical limit is imposed here.

[0183] S105. Obtain historical temperature data and historical vibration data. Calculate the historical vibration data variance, historical vibration eigenvalue, historical temperature data variance, historical temperature eigenvalue, and historical spatiotemporal correlation eigenvalue based on the historical temperature data and historical vibration data, and form an eigenvector.

[0184] For example, the calculation methods for historical vibration characteristic values, historical temperature characteristic values, and historical spatiotemporal correlation characteristic values ​​can refer to the calculation methods in steps S101-S104, and will not be repeated here.

[0185] Furthermore, the variance of historical temperature data can be obtained by taking the maximum value after obtaining the variance of each historical comparative temperature data, or by calculating the average value after obtaining the variance of each historical comparative temperature data. Either of these two implementation methods is acceptable, and either one can be chosen.

[0186] After calculating the historical vibration data variance, historical vibration eigenvalues, historical temperature data variance, historical temperature eigenvalues, and historical spatiotemporal correlation eigenvalues, they can be combined into an eigenvector. The eigenvector can take the form of [historical vibration data variance, historical vibration eigenvalues, historical temperature data variance, historical temperature eigenvalues, historical spatiotemporal correlation eigenvalues].

[0187] Alternatively, to increase the amount of feature vector data, a moving window can be used to obtain the feature vectors:

[0188] For example, a 5-minute window can be used to slide across historical data, with the sliding step size set to either 5 minutes or 3 minutes. Each window's historical data can be considered as a set of historical temperature and vibration data, and a corresponding feature vector can be calculated. This calculation can generate multiple windows of historical data, forming multiple feature vectors.

[0189] S106. Train the random forest model by using the feature vector as the input and the data label corresponding to the feature vector as the output of the random forest model to obtain the trained random forest model, where the data label is the binary label of the feature vector.

[0190] For example, the data labels corresponding to the feature vectors can be set using a machine learning model and can be manually corrected. The data labels are binary labels, including 0: normal, no aggregation and 1: particle aggregation exists.

[0191] S107. The variance of vibration data, vibration eigenvalues, variance of temperature data, temperature eigenvalues, and spatiotemporal correlation eigenvalues ​​are combined to form the vector to be analyzed. The vector to be analyzed is input into the trained random forest model. The trained random forest model outputs data labels. The speed of the mixer is adjusted by the data labels.

[0192] For example, the method for calculating the variance of the temperature data in this embodiment must be consistent with the method for calculating the variance of historical temperature data. If the method for calculating the variance of historical temperature data is to obtain the maximum value after obtaining the variance of each historical comparative temperature data, then the variance of the temperature data here also needs to be the maximum value of the variance of each comparative temperature data. If the method for calculating the variance of historical temperature data is to calculate the average value after obtaining the variance of each historical comparative temperature data, then the variance of the temperature data here also needs to be the average value of the variance of each comparative temperature data. This needs to be consistent with the calculation method used when calculating the feature vector of historical temperature data.

[0193] State determination and instruction generation: The vector to be analyzed, calculated in real time (composed of the variance of vibration data, vibration eigenvalues, variance of temperature data, temperature eigenvalues, and spatiotemporal correlation eigenvalues), is input into the trained random forest model. The model outputs the data label (binary output) corresponding to this vector.

[0194] If the output label is 0, it indicates that the current mixing state is uniform and there is no significant agglomeration, and the control system continues to operate with the current parameters. If the output label is 1, it confirms the presence of anti-slip particle agglomerates. The control system then generates targeted control commands, such as "increase the agitator speed".

[0195] Command execution and mechanism: The control system sends an acceleration command to the variable frequency drive of the mixing motor. The increased motor speed drives the mixing paddle (or roller) to rotate faster, thereby applying greater shear force to the mixture. This enhanced shearing action effectively targets anomalous areas implied by spatiotemporal correlation characteristics, breaking up located agglomerates of anti-slip particles.

[0196] Effect Verification and Closed-Loop Adjustment: After a period of accelerated operation, the system continues to acquire RTD temperature and vibration data in real time and repeats the feature extraction and model analysis process described above. If the previously detected periodic thermal anomalies (manifested as an increase in temperature characteristic values) and impact vibrations (manifested as an increase in vibration characteristic values) weaken or disappear synchronously, and the data labels output by the model return to 0, it indicates that the accelerated processing is effective and particle agglomeration has been eliminated. At this point, the system can control the rotation speed to gradually return to the normal process setpoint.

[0197] The rubber base material and anti-slip particles can then be thoroughly mixed on an open mill to form a micro-uniformly distributed blend. Next, the blend is calendered into a slab: the blend is rolled through a calender to form a continuous slab with a thickness slightly greater than the final product thickness.

[0198] Cooling and shaping: The slab is cooled to room temperature in the cooling section to stabilize its size and shape.

[0199] The precision slicing process can include: fixed-length cutting: cutting the cooled slab to a set length; and thickness slicing: using high-precision longitudinal cutting or slicing equipment to cut the slab into multiple thin slices of equal thickness along the thickness direction. Each slice serves as a pre-fabricated unit for subsequent vulcanization, with its internal anti-slip particles evenly distributed in three dimensions.

[0200] The compression molding and vulcanization process can include: Stacking / Single-sheet molding: Depending on the product thickness requirements, one or more pre-made sheets are stacked in a flat vulcanization mold. Hot-press vulcanization: Compression vulcanization is performed under set temperature, pressure, and time. Under the action of heat and pressure, the rubber cross-links, transforming from a linear structure to a three-dimensional network structure, achieving final elasticity and strength. Simultaneously, the layers (if stacked) fuse together to form the final outline and surface morphology of the product. Cooling and demolding: After vulcanization, the mold is cooled, and the molded rubber flooring is removed.

[0201] Post-processing and inspection may include: Trimming: Removing burrs and flash from the product's perimeter. Surface cleaning: Cleaning the product surface. Quality inspection: Conducting full or random inspections of the finished flooring's dimensions, thickness, hardness, slip resistance, and appearance. Packaging and warehousing: Packaging qualified products and transferring them to the finished goods warehouse.

[0202] In summary, in this embodiment of the invention, by using multimodal sensor data fusion and spatiotemporal correlation analysis, early and accurate identification and active closed-loop control of particle agglomeration during the mixing process are achieved, thereby significantly improving the consistency of anti-slip performance and quality stability of the finished rubber flooring, while reducing the scrap rate and production costs.

[0203] This invention also proposes an artificial intelligence-based system for controlling the mixing uniformity of granules in sheet rubber flooring. Please refer to [link / reference]. Figure 2The diagram shows a structural diagram of an artificial intelligence-based particle mixing uniformity control system for polished rubber flooring provided in an embodiment of the present invention. The system includes: a data acquisition module 101, a data processing module 102, and a speed adjustment module 103.

[0204] The data acquisition module 101 is used to acquire temperature data and vibration data during the mixing process of the granulated rubber flooring particles. The temperature data includes reference temperature data and at least two comparison temperature data.

[0205] The data processing module 102 is used to determine the period of vibration data, divide the vibration data according to the period to obtain period segments, and calculate the vibration characteristic value of the vibration data based on the vibration data in each period segment; determine the temperature change rate sequence based on the reference temperature data, and calculate the temperature characteristic value of the temperature data based on the temperature change rate sequence and at least two comparative temperature data; obtain the spatiotemporal correlation characteristic value of the temperature data and vibration data; obtain historical temperature data and historical vibration data, and calculate the historical vibration data variance, historical vibration characteristic value, historical temperature data variance, historical temperature characteristic value and historical spatiotemporal correlation characteristic value based on the historical temperature data and historical vibration data, and form a feature vector; use the feature vector as the input of the random forest model and the data label corresponding to the feature vector as the output of the random forest model to train the random forest model, and obtain the trained random forest model, where the data label is the binary label of the feature vector;

[0206] The speed adjustment module 103 is used to form an analysis vector from the variance of vibration data, vibration feature value, variance of temperature data, temperature feature value and spatiotemporal correlation feature value, input the analysis vector into a trained random forest model, output data labels through the trained random forest model, and adjust the speed of the mixer through the data labels.

[0207] It should be noted that the system provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the artificial intelligence-based particle mixing uniformity control system for polished rubber flooring provided in the above embodiments and the artificial intelligence-based particle mixing uniformity control method for polished rubber flooring provided in the above embodiments belong to the same concept. The specific implementation process is detailed in the method embodiments and will not be repeated here.

[0208] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0209] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0210] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for controlling the mixing uniformity of granules in polished rubber flooring based on artificial intelligence, characterized in that, include: Acquire temperature and vibration data during the mixing process of the granulated rubber flooring particles, wherein the temperature data includes baseline temperature data and at least two comparative temperature data; Determine the period of the vibration data, divide the vibration data according to the period to obtain period segments, and calculate the vibration characteristic value of the vibration data based on the vibration data in each period segment; Determine the temperature change rate sequence based on the reference temperature data, and calculate the temperature characteristic value of the temperature data based on the temperature change rate sequence and at least two comparative temperature data. Acquire the spatiotemporal correlation feature values ​​of temperature and vibration data; Acquire historical temperature data and historical vibration data, calculate historical vibration data variance, historical vibration eigenvalues, historical temperature data variance, historical temperature eigenvalues, and historical spatiotemporal correlation eigenvalues ​​based on the historical temperature data and historical vibration data, and form an eigenvector; The random forest model is trained by using the feature vector as input and the corresponding data label as output. The data label is the binary label of the feature vector. The variance of vibration data, vibration eigenvalues, variance of temperature data, temperature eigenvalues, and spatiotemporal correlation eigenvalues ​​are used to form an analysis vector. This analysis vector is then input into a trained random forest model, which outputs data labels. These data labels are used to adjust the mixer's rotation speed.

2. The method for controlling the mixing uniformity of polished rubber flooring particles based on artificial intelligence according to claim 1, characterized in that, The determination of the period of the vibration data specifically includes: Perform a Fourier transform on the vibration data to obtain the Fourier transform result; The maximum amplitude is obtained from the Fourier transform result, and the reciprocal of the maximum amplitude is used to determine the period. 3.The AI-based sectioning rubber floor particle uniformity regulation method of claim 1, wherein, The calculation of vibration characteristic values ​​based on vibration data in each period segment specifically includes: The vibration data in each period segment are clustered separately to obtain the vibration data category for each period segment; For each period segment, the maximum vibration value in the vibration data category with the largest amount of data is determined as the basic number of the period segment; The difference between each vibration value and the basic number in the period segment is obtained, and the ratio of the difference to the basic number is used to determine the vibration anomaly of each vibration value. Vibration values ​​within a periodic segment that exhibit abnormal vibration exceeding a preset abnormal vibration threshold are defined as abnormal vibration values. For the entire periodic segment, the time point of each abnormal vibration value is determined as the abnormal time point; Obtain the ratio of the frequency of each abnormal time point to the number of periodic segments, and determine the maximum ratio as the vibration characteristic value.

4. The AI-based sectioning rubber floor particle uniformity regulation method of claim 3, wherein, The time point for each abnormal vibration value is obtained as follows: The difference between the time corresponding to the nth abnormal vibration value in the mth period segment and the time corresponding to the first vibration value in the mth period segment is determined as the time point of the nth abnormal vibration value in the mth period segment. Obtain the time point for each abnormal vibration value in the m-th period segment; Obtain the time point for each abnormal vibration value in each period segment. 5.The AI-based sectioning rubber floor particle uniformity regulation method of claim 1, wherein, The step of determining the temperature change rate sequence based on reference temperature data, and calculating the temperature characteristic value of the temperature data based on the temperature change rate sequence and at least two comparative temperature data, specifically includes: The absolute value of the difference between the a-th temperature value and the (a-1)-th temperature value in the reference temperature data is taken, and the ratio of the absolute value to the (a-1)-th temperature value is determined as the temperature change rate of the a-th temperature value. The a-th temperature value is not the first temperature value in the reference temperature data. The sequence of temperature change rates for each temperature value in the reference temperature data is defined as the temperature change rate sequence. The maximum value in the temperature change rate sequence is determined as the reference temperature change rate; Obtain the temperature change rate sequence for each comparative temperature data point; In each comparison temperature change rate sequence, elements that are greater than the baseline temperature change rate are identified as outlier values. The temperature value corresponding to the abnormal element value is identified as the abnormal temperature value; For each abnormal temperature value, calculate the difference between the comparative temperature change rate and the baseline temperature change rate, and determine the temperature abnormality of each abnormal temperature value by the ratio of the difference to the baseline temperature change rate. The average of all abnormal temperature values ​​is used to obtain the temperature characteristic value.

6. The AI-based sectioning rubber floor particle uniformity regulation method of claim 5, wherein, The acquisition of spatiotemporal correlation feature values ​​of temperature data and vibration data specifically includes: The mean value of temperature values ​​in the baseline temperature data is determined as the normal temperature value, and the ratio of abnormal temperature values ​​to normal temperature values ​​is determined as the temperature anomaly rate. In the comparison temperature data, the abnormal temperature value in the first comparison temperature data is determined as the first abnormal temperature data, and the abnormal temperature value in the second comparison temperature data is determined as the second abnormal temperature data. The temperature anomaly rate of each first abnormal temperature data is determined as the first temperature anomaly rate, and the temperature anomaly rate of each second abnormal temperature data is determined as the second temperature anomaly rate. The first temperature anomaly rate and the second temperature anomaly rate are respectively determined as the matching set of the extended KM matching algorithm, and the ratio of the temperature anomaly rates is determined as the weight of the edge of the extended KM matching algorithm. The extended KM matching algorithm is run to obtain the matching relationship. The first temperature anomaly rate and the second temperature anomaly rate are clustered separately to obtain the first cluster set and the second cluster set; Based on the matching relationship, determine the matching relationship between the first cluster set and the second cluster set, and based on the matching relationship between the first cluster set and the second cluster set, determine the matching cluster pairs, and determine the set of matching cluster pairs as the matching set; For each matching cluster pair in the matching set, obtain the first and second comparative temperature data of the matching cluster pair; The time point in the comparative temperature data for each target temperature anomaly rate is determined as the target anomaly time point, where the target temperature anomaly rate is the first temperature anomaly rate of the first comparative temperature data and the second temperature anomaly rate of the second comparative temperature data. Calculate the spatiotemporal correlation feature value based on the target anomaly time point.

7. The method for controlling the mixing uniformity of polished rubber flooring particles based on artificial intelligence according to claim 6, characterized in that, The process of determining the matching relationship between the first cluster set and the second cluster set based on the matching relationship, determining matching cluster pairs based on the matching relationship between the first cluster set and the second cluster set, and determining the set of matching cluster pairs as the matching set specifically includes: Based on the matching relationship, the second temperature anomaly rate corresponding to the cth first temperature anomaly rate in the bth cluster of the first cluster set is determined as the matching element, and the clusters of the matching element in the second cluster set are determined as candidate matching clusters. Obtain the candidate matching clusters for the b-th cluster in the first set of clusters; The ratio of the number of matching elements in the candidate matching cluster to the number of elements in the candidate matching cluster is determined as the matching value of the candidate matching cluster. The candidate matching clusters whose matching value is greater than the preset matching threshold are determined as the matching clusters of the b-th cluster in the first cluster set, and the b-th cluster in the first cluster set and the matching cluster are determined as a matching cluster pair; Any data point in the comparative temperature data is designated as the first comparative temperature data point, and any data point other than the first comparative temperature data point is designated as the second comparative temperature data point. Matching cluster pairs are obtained, and the set of matching cluster pairs is determined as the matching set.

8. The AI-based sectioning rubber floor particle mixture uniformity regulation method of claim 6, wherein, The calculation of spatiotemporal correlation feature values ​​based on the target anomaly time point specifically includes: The time points in the vibration data corresponding to the abnormal time points where the vibration characteristic value is greater than the preset vibration characteristic threshold are determined as abnormal vibration time points; Obtain the target time point of the anomaly; Calculate the absolute value of the difference between the z-th target anomaly time point and each vibration anomaly time point to obtain the matching duration set of the z-th target anomaly time point; The vibration anomaly time point corresponding to the element in the matching time set that is less than the preset time threshold is determined as the matching time point of the z-th target anomaly time point; Obtain the matching time point for each target anomaly time point; Obtain the ratio of the number of matching time points to the number of vibration anomaly time points, and determine the maximum ratio as the spatiotemporal correlation feature value. 9.The AI-based sectioning rubber floor particle uniformity regulation method of claim 6, wherein, The first temperature anomaly rate and the second temperature anomaly rate are obtained by sorting the temperature anomaly rates in ascending order. When the first temperature anomaly rate is greater than the second temperature anomaly rate, the temperature anomaly rate is the ratio of the second temperature anomaly rate to the first temperature anomaly rate. When the first temperature anomaly rate is less than or equal to the second temperature anomaly rate, the temperature anomaly rate is the ratio of the first temperature anomaly rate to the second temperature anomaly rate. 10.A system for regulating the uniformity of rubber floor particles based on artificial intelligence, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that, When the computer program is executed by the processor, it implements the steps of the artificial intelligence-based method for controlling the mixing uniformity of granules in sheet rubber flooring as described in any one of claims 1-9.

Images