A dynamic data intelligent control system and method

Abstract

The application belongs to the technical field of intelligent device control, and discloses a dynamic data intelligent control system and method; comprising: collecting multi-dimensional running state data of a moving object in real time; collecting historical multi-dimensional running state data and calculating the entropy value of the moving object; preliminarily evaluating whether the moving object has resistance abnormality according to the entropy value of the moving object; if there is preliminary resistance abnormality, extracting the dynamic change rate index of the moving object; outputting an evaluation result based on the multi-dimensional running state data of the moving object, the dynamic change rate index and a preset resistance evaluation model, wherein the evaluation result comprises resistance normality or resistance abnormality; if it is determined that there is resistance abnormality, generating a resistance regulation strategy through a group intelligence optimization algorithm according to the multi-dimensional running state data of the moving object, executing and feeding back data; the application can realize quantitative and accurate friction reduction of the moving object and reduce the failure risk of the moving object.

Description

Technical Field

[0001] This invention relates to the field of intelligent device control technology, and more specifically, to a dynamic data intelligent control system and method. Background Technology

[0002] Moving objects, as key basic equipment in the industrial field, have wide and important applications in industries such as mining, building materials, and chemicals. Their working efficiency and operational stability are directly related to the quality and output of the production process. However, during long-term operation, moving objects are affected by factors such as vibration and friction, making the effectiveness of the friction reduction system a crucial link in ensuring the normal operation of the equipment. Traditional friction reduction methods rely heavily on manual experience, which has revealed many drawbacks in practical applications: First, they are prone to delayed leakage; too much or too little friction-reducing oil supply will adversely affect the service life and working efficiency of the equipment. Second, they cannot monitor the quality of friction reduction and the wear and tear of components in real time, making it difficult to quickly identify potential problems. Third, they lack the ability to optimize resistance-related parameters, resulting in a bottleneck in improving equipment efficiency.

[0003] While existing technologies can control the resistance of moving objects, they do not detail how to control resistance based on multiple collected parameters. They lack targeted resistance control methods and struggle to effectively correlate multiple parameters such as vibration and temperature with resistance-related parameters. This results in the inability to effectively integrate and utilize the collected data, affecting the real-time adjustment and accuracy of the resistance control strategy. Furthermore, the aforementioned technologies only briefly describe injecting a certain amount of grease into the bearing without explaining the specific control logic and implementation methods. They lack in-depth analysis and explanation of the friction reduction process, such as how to determine the resistance-related parameters like the amount and temperature of the medium oil. This leads to improper control of resistance-related parameters, increasing wear and failure risks of moving objects, shortening their service life, and ultimately affecting production efficiency and safety.

[0004] In view of this, the present invention proposes a dynamic data intelligent control system and method to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art and achieve the above objectives, the present invention provides the following technical solution: a dynamic data intelligent control method, comprising:

[0006] Real-time acquisition of multi-dimensional operational status data of moving objects;

[0007] Collect historical multidimensional operational status data and calculate the entropy value of moving objects;

[0008] Based on the entropy value of the moving object, a preliminary assessment can be made as to whether the moving object exhibits abnormal resistance.

[0009] If there is an initial resistance anomaly, extract the dynamic rate of change index of the moving object;

[0010] Based on the multidimensional operating status data of the moving object, the dynamic rate of change index, and the preset lubrication evaluation model and resistance evaluation model, the evaluation results are output, including whether the resistance is normal or abnormal.

[0011] If the resistance is determined to be abnormal, a resistance control strategy is generated based on the multi-dimensional operational status data of the moving object, and then executed and data is fed back.

[0012] The resistance-related parameters are dynamically adjusted based on the feedback data, and the resistance control strategy is redefined. The feedback data consists of multi-dimensional operating status data after the resistance control is completed.

[0013] Furthermore, methods for generating resistance control strategies using swarm intelligence optimization algorithms include:

[0014] Step 1: Construct n sets of resistance-related parameters, where n is an integer greater than 1. Each set contains a random combination of moving part medium temperature, working medium delivery pressure, and working medium flow rate. Set different parameter labels for each set of resistance-related parameters.

[0015] Step 2: Initialize the population, define the correspondence between the individual positions and parameter labels in the one-dimensional search space, and set the initial number of iterations to 0;

[0016] Step 3: Calculate the fitness function, predict screening quality and screening efficiency using a deep belief network model, and determine the fitness function based on a weighted sum;

[0017] Step 4: Iteratively update the population, including:

[0018] Update the location of explored and followed individuals to generate an updated population;

[0019] Apply a Gaussian perturbation to individuals in the updated population to generate a perturbed population;

[0020] Calculate the dynamic boundary, update the reverse solutions of individuals in the population, and generate a reverse solution population;

[0021] After merging populations, the optimal set of resistance-related parameters is selected based on the selection probability until the iteration threshold is reached. The set of resistance-related parameters corresponding to the individual with the highest fitness is then used as the resistance control strategy.

[0022] Furthermore, the method for obtaining screening quality and screening efficiency is as follows: the set of resistance-related parameters corresponding to the test data and the individual position is used as analysis data. The test data includes multi-dimensional operating status data and dynamic change rate indicators of moving objects. Based on the analysis data and the preset quality prediction model and efficiency prediction model, the corresponding screening quality and screening efficiency are predicted. Both the quality prediction model and the efficiency prediction model are deep belief network models.

[0023] Furthermore, the first individual in the population is the exploratory individual, and the remaining individuals are the follower individuals; the range of parameter labels and the range of the one-dimensional search space are both [1, n].

[0024] Furthermore, methods for setting population size include:

[0025] Population size ranges are defined based on historical test data; population size ranges are dynamically adjusted by segmentation ratios, and fitness comparison results are calculated.

[0026] Based on the fitness comparison results, the population size range is narrowed until the range width is less than the threshold. The average of the maximum and minimum values ​​in the updated population size range is then taken as the population size.

[0027] Furthermore, updated methods for exploring individual locations include:

[0028] The exploration individuals, second position, and last position are determined by fitness ranking;

[0029] The position update of the explorer is based on the random weighted offset between the position of the individual with the highest fitness and the second or last position in the population. The offset direction is determined by a random threshold, and the offset is dynamically adjusted by combining the random perturbation amplitude of the number of iterations to update the position of the explorer.

[0030] The position update of an individual is generated by taking the average of its current position and the position of the previous individual.

[0031] Furthermore, the method for calculating the entropy value of a moving object includes:

[0032] Historical multidimensional operational status data and identical data from the multidimensional operational status data are added to an analysis set, ensuring a one-to-one correspondence between the data in the analysis set and the data in the multidimensional operational status data. A clustering algorithm is used to cluster the data in each analysis set, obtaining each cluster corresponding to each analysis set. The number of data points in each cluster is obtained, and the numbers of each data point in each analysis set are summed sequentially to obtain the total number of data points in each analysis set. The number of data points in each cluster is divided by the total number of data points in the corresponding analysis set to obtain the interval probability of each cluster. Based on the interval probability of each cluster in each analysis set, the set entropy value of each analysis set is calculated. The set entropy value of each analysis set is used as the entropy value of the moving object.

[0033] Furthermore, the method for initially assessing whether a moving object exhibits abnormal drag includes:

[0034] The process involves: acquiring historical entropy values ​​(calculated at historical moments); marking real-time calculated entropy values ​​as real-time entropy values; summing the entropy values ​​of each set in the real-time entropy values ​​with the corresponding set in the historical entropy values ​​to obtain the total entropy value of the moving object for each data point in the multidimensional operational state data; counting the number of moving object entropy values ​​in the historical entropy values ​​and adding one to obtain the entropy count; dividing the total entropy value of each vibration by the entropy count to obtain the mean entropy value for each data point in the multidimensional operational state data; subtracting the mean entropy value from each set of entropy values ​​in the real-time and historical entropy values ​​and squaring the result to obtain the entropy difference; summing the entropy differences for each data point in the multidimensional operational state data and dividing by the entropy count to obtain the variance for each data point in the multidimensional operational state data; taking the square root of each variance to obtain the standard deviation of the entropy value for each data point in the multidimensional operational state data; and adding three times the standard deviation of the entropy value to the mean entropy value for each data point in the multidimensional operational state data to obtain the entropy threshold.

[0035] The entropy value of each set in the real-time entropy values ​​is compared with the corresponding entropy threshold.

[0036] If the entropy value of each set is less than or equal to the corresponding entropy threshold, then the preliminary assessment is that the moving object does not have any resistance anomaly.

[0037] If the entropy value of a set is greater than the corresponding entropy threshold, it is preliminarily assessed that the moving object has an abnormal resistance.

[0038] Furthermore, the method for re-formulating the resistance control strategy includes:

[0039] The multi-dimensional operational status data collected in real time is marked as real-time data;

[0040] The entropy of the moving object is calculated based on the feedback data, and the presence of resistance anomalies is assessed. If no resistance anomalies are found, the resistance control strategy is not revised. If resistance anomalies still exist, the set of resistance-related parameters in the resistance control strategy and real-time data are input into the trained parameter prediction model to predict the corresponding data. The predicted data includes vibration frequency, vibration amplitude, bearing temperature, and screen temperature. The frequency prediction model predicts the vibration frequency, the amplitude prediction model predicts the vibration amplitude, the bearing temperature prediction model predicts the bearing temperature, and the screen temperature prediction model predicts the screen temperature. The parameter prediction model includes a frequency prediction model, an amplitude prediction model, a bearing temperature prediction model, and a screen temperature prediction model. Each model in the parameter prediction model is a deep belief network model. Each data point in the predicted data is subtracted from the corresponding data point in the feedback data to obtain the unadjusted amount for each data point. Each data point in the predicted data is subtracted from the corresponding data point in the real-time data to obtain the amount to be adjusted for each data point. The unadjusted amount for each data point in the predicted data is divided by the amount to be adjusted to calculate the unadjusted rate for each data point in the predicted data. Each unadjusted rate is incremented by 1 to obtain the adjustment rate for each data point in the predicted data.

[0041] The vibration frequency, vibration amplitude, bearing temperature, and screen temperature in the real-time data are multiplied by the corresponding adjustment rate to obtain the actual data; the screen load in the feedback data is also added to the actual data, and the dynamic change rate index is extracted from the actual data and marked as the actual characteristic data; based on the actual data and the actual characteristic data, the resistance control strategy is re-formulated.

[0042] A dynamic data intelligent control system, implementing the aforementioned dynamic data intelligent control method, includes:

[0043] The data acquisition module is used to collect multi-dimensional operational status data of moving objects in real time;

[0044] The data analysis module is used to collect historical multidimensional operational status data and calculate the entropy value of moving objects;

[0045] The preliminary assessment module uses the entropy value of the moving object to make a preliminary assessment of whether the moving object has abnormal resistance.

[0046] The feature extraction module extracts the dynamic rate of change index of the moving object if there is an initial resistance anomaly.

[0047] The anomaly assessment module outputs assessment results based on the multidimensional operational status data of the moving object, the dynamic rate of change index, and the preset resistance assessment model. The assessment results include whether the resistance is normal or abnormal.

[0048] If the resistance control module determines that the resistance is abnormal, it generates a resistance control strategy based on the multi-dimensional operating state data of the moving object through a swarm intelligence optimization algorithm, executes the strategy, and feeds back the data.

[0049] The feedback optimization module dynamically adjusts the resistance-related parameters based on the feedback data and re-formulates the resistance control strategy. The feedback data consists of multi-dimensional operating status data after the resistance control is completed.

[0050] The technical effects and advantages of the dynamic data intelligent control system and method of the present invention are as follows:

[0051] By comprehensively monitoring the operating status of moving objects through multi-parameter acquisition, entropy analysis is used to initially assess the resistance status; feature data is extracted and combined with deep learning technology to further assess the resistance status and promptly identify abnormal resistance conditions of moving objects; a swarm intelligence optimization algorithm is used to formulate a precise resistance control strategy and perform dynamic feedback adjustment; it can effectively integrate and utilize multiple collected data and perform accurate calculation of resistance-related parameters to achieve quantitative and precise friction reduction of moving objects, improve the operating efficiency and service life of moving objects, and reduce the risk of failure of moving objects. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of a dynamic data intelligent control system according to Embodiment 1 of the present invention;

[0053] Figure 2 A flowchart illustrating the steps for formulating the resistance control strategy in Embodiment 1 of the present invention;

[0054] Figure 3 This is a flowchart of the iterative population update steps in Embodiment 1 of the present invention;

[0055] Figure 4 This is a flowchart of a dynamic data intelligent control method according to Embodiment 2 of the present invention. Detailed Implementation

[0056] 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.

[0057] Example 1

[0058] Please see Figure 1As shown in the figure, the dynamic data intelligent control system described in this embodiment includes a data acquisition module, a data analysis module, a preliminary evaluation module, a feature extraction module, an anomaly evaluation module, a resistance control module, and a feedback optimization module; the modules are connected by wired and / or wireless means to realize data transmission between the modules;

[0059] The data acquisition module is used to collect multi-dimensional operating status data of moving objects in real time. The moving objects include, but are not limited to, vibrating screens, motors, compressors, etc. The factors affecting their resistance include, but are not limited to, the lubrication effect of the medium oil (such as lubricating oil) during movement.

[0060] The multi-dimensional operating status data includes vibration frequency, vibration amplitude, bearing temperature, screen temperature, and screen load. Vibration frequency is the number of vibrations the screen undergoes per unit time in the moving object, affecting screening efficiency and material handling capacity; it is obtained by a vibration sensor installed on the screen. Vibration amplitude is the displacement range of the screen during vibration, affecting screening effect and material flow; it is obtained by a displacement sensor installed on the screen. Bearing temperature is the operating temperature of the bearing in the moving object; excessively high temperatures may lead to medium oil failure and bearing damage; it is obtained by a temperature sensor installed on the bearing housing. Screen temperature is the operating temperature of the screen in the moving object, affecting material flowability and screening effect; it is obtained by a temperature sensor installed on the screen. Screen load is the weight of material applied to the screen, affecting screening efficiency and screen lifespan; it is obtained by a pressure sensor installed on the support structure in the moving object. The multi-dimensional operating status data is collected according to a preset sampling frequency, which is preset by those skilled in the art based on actual conditions.

[0061] It should be noted that the reason for collecting multi-dimensional operational status data is as follows: monitoring vibration frequency can assess the operating status and workload of moving objects; abnormal changes in vibration frequency lead to increased friction, thus requiring adjustment of the resistance control strategy; vibration amplitude reflects the vibration intensity and stability of moving objects; large vibration amplitudes can cause uneven flow of the medium oil, thus requiring adjustment of the resistance control strategy; bearing temperature is an important indicator for judging the resistance status; excessively high bearing temperatures usually indicate insufficient lubrication or a decline in the quality of the medium oil, thus requiring adjustment of the resistance control strategy; screen temperature directly affects the flowability and screening effect of materials; excessively high screen temperature can lead to material adhesion and blockage; as the screen temperature increases, the performance of the medium oil decreases, thus requiring adjustment of the resistance control strategy; screen load directly affects screening efficiency and the stability of moving objects; when the screen load increases, the supply of medium oil needs to be increased to cope with greater friction and wear risks; by collecting vibration characteristic data, it is helpful to subsequently optimize the resistance control strategy, adjust resistance correlation parameters, achieve more precise and intelligent resistance control, and improve the operating efficiency and reliability of moving objects;

[0062] The data analysis module is used to collect historical multidimensional operational status data and calculate the entropy value of moving objects;

[0063] Historical multidimensional operational status data refers to multidimensional operational status data collected at historical moments.

[0064] Methods for calculating the entropy of a moving object include:

[0065] Historical multidimensional operational status data and identical data from the multidimensional operational status data are added to an analysis set, meaning there is a one-to-one correspondence between the analysis set and the data in the multidimensional operational status data. Clustering algorithms (such as K-Means clustering, hierarchical clustering, DBSCAN, etc.) are used to cluster the data in each analysis set, obtaining each cluster corresponding to each analysis set. The number of data points in each cluster is obtained, and the number of data points for each analysis set is summed sequentially to obtain the total number of data points in each analysis set. The number of data points in each cluster is divided by the total number of data points in the corresponding analysis set to obtain the interval probability of each cluster. Based on the interval probability of each cluster in each analysis set, the set entropy value of each analysis set is calculated. The set entropy value of each analysis set is used as the entropy value of the moving object.

[0066] The expression for the entropy of a set is: In the formula, H is the set entropy value, p j Let J be the interval probability of the j-th cluster, where j∈[1,J] and J is the total number of clusters;

[0067] The preliminary assessment module uses the entropy value of the moving object to make a preliminary assessment of whether the moving object has abnormal resistance.

[0068] Methods for preliminary assessment of whether a moving object exhibits abnormal drag include:

[0069] The process involves: acquiring historical entropy values ​​(calculated at historical moments); marking real-time calculated entropy values ​​as real-time entropy values; summing the entropy values ​​of each set in the real-time entropy values ​​with the corresponding set in the historical entropy values ​​to obtain the total entropy value of the moving object for each data point in the multidimensional operational state data; counting the number of moving object entropy values ​​in the historical entropy values ​​and adding one to obtain the entropy count; dividing the total entropy value of each vibration by the entropy count to obtain the mean entropy value for each data point in the multidimensional operational state data; subtracting the mean entropy value from each set of entropy values ​​in the real-time and historical entropy values ​​and squaring the result to obtain the entropy difference; summing the entropy differences for each data point in the multidimensional operational state data and dividing by the entropy count to obtain the variance for each data point in the multidimensional operational state data; taking the square root of each variance to obtain the standard deviation of the entropy value for each data point in the multidimensional operational state data; and adding three times the standard deviation of the entropy value to the mean entropy value for each data point in the multidimensional operational state data to obtain the entropy threshold.

[0070] The entropy value of each set in the real-time entropy values ​​is compared with the corresponding entropy threshold.

[0071] If the entropy value of each set is less than or equal to the corresponding entropy threshold, then the preliminary assessment is that the moving object does not have any resistance anomaly.

[0072] If there is a set entropy value greater than the corresponding entropy threshold, it is preliminarily assessed that the moving object has an abnormal resistance.

[0073] The feature extraction module extracts the dynamic rate of change index of the moving object if there is an initial resistance anomaly.

[0074] Methods for extracting the rate of change index of a moving object include:

[0075] Historical multidimensional operational status data is acquired, which is the multidimensional operational status data of the moving object collected in the last time. The vibration frequency, vibration amplitude, bearing temperature, and screen temperature in the multidimensional operational status data of the moving object are subtracted from the vibration frequency, vibration amplitude, bearing temperature, and screen temperature in the historical multidimensional operational status data, respectively, to obtain the changes in vibration frequency, vibration amplitude, bearing temperature, and screen temperature. The changes in vibration frequency, vibration amplitude, bearing temperature, and screen temperature are then divided by the sampling frequency to obtain the rate of change of vibration frequency, rate of change of vibration amplitude, rate of change of bearing temperature, and rate of change of screen temperature, which are used as the dynamic rate of change indicators of the moving object.

[0076] The anomaly assessment module outputs assessment results based on the multidimensional operational status data of the moving object, the dynamic rate of change index, and the preset resistance assessment model. The assessment results include whether the resistance is normal or abnormal.

[0077] Methods for assessing whether a moving object exhibits abnormal drag include:

[0078] Multidimensional operational status data and dynamic rate of change indicators are used as test data. The test data is input into the trained resistance assessment model, which outputs assessment labels. The corresponding assessment results are obtained based on the assessment labels, and the moving object is assessed for whether there is an abnormal resistance. The assessment label is a numerical label corresponding to the assessment result. The assessment results include normal and abnormal. Different assessment results correspond to different numerical labels. For example, the numerical label for normal is set to 1, and the numerical label for abnormal is set to 2.

[0079] The specific training process of the resistance assessment model includes:

[0080] Pre-set corresponding judgment results for group a of test data, where a is an integer greater than 1. The judgment results corresponding to the test data are obtained by those skilled in the art during the historical process of controlling the resistance of moving objects. Different test data of group a are collected. Based on practical experience, those skilled in the art analyze different test data of group a in sequence to determine whether the moving object has abnormal resistance. The corresponding judgment results are set for different test data of group a in sequence.

[0081] The test data and corresponding evaluation labels are converted into a set of feature vectors. Each set of feature vectors is used as input to the resistance assessment model, which outputs a set of predicted evaluation labels corresponding to each set of test data and uses the actual evaluation labels corresponding to each set of test data as the prediction target. The actual evaluation labels are the pre-collected evaluation labels corresponding to the test data. The mean absolute percentage error (MAPE) is used to evaluate the model accuracy of the prediction results. When the calculated MAPE is less than the preset MAPE, the resistance assessment model training is complete. The formula for calculating MAPE is as follows: Where, η b The actual evaluation label corresponding to the test data of group b. The predicted evaluation label is given for the b-th group of test data, where b is the group number of the feature vector corresponding to the test data, and b∈[1,a]; a resistance evaluation model is generated based on the predicted evaluation label of the test data; wherein, the resistance evaluation model is a deep belief network model, and the preset MAPE is preset by those skilled in the art according to the required accuracy of the resistance evaluation model.

[0082] It should be noted that the reason for first calculating the entropy value and entropy threshold of the moving object for preliminary resistance anomaly assessment, and then calculating the dynamic change rate index and combining it with deep learning technology for final resistance anomaly assessment, is that: entropy value can quickly reflect the complexity and uncertainty of data, helping to initially determine whether there is resistance anomaly in the moving object; if the entropy value increases, it may indicate poor resistance status; the change rate can capture the trend of data changes over time, helping to identify whether the friction reduction supply is appropriate; and the change rate reflects real-time, subtle changes, making it more sensitive to anomaly judgment; deep learning technology can process high-dimensional data, automatically extract features, and reduce the need for manual feature engineering; it is also good at capturing nonlinear relationships in data, and can better identify the complex relationship between resistance status and moving object characteristics; in addition, deep learning models can adaptively adjust through training, and with more data input, the accuracy of the model can be continuously improved, adapting to new working conditions or environmental changes.

[0083] If the resistance control module determines that the resistance is abnormal, it generates a resistance control strategy based on the multi-dimensional operating state data of the moving object through a swarm intelligence optimization algorithm, executes the strategy, and feeds back the data.

[0084] like Figure 2 As shown, the method for generating resistance control strategies using swarm intelligence optimization algorithms includes:

[0085] Step 1: Construct n sets of resistance-related parameters, where n is an integer greater than 1. Each set contains a random combination of moving part medium temperature, working medium delivery pressure, and working medium flow rate. Set different parameter labels for each set of resistance-related parameters.

[0086] Step 2: Initialize the population, define the correspondence between the individual positions and parameter labels in the one-dimensional search space, and set the initial number of iterations to 0;

[0087] Step 3: Calculate the fitness function, predict screening quality and screening efficiency using a deep belief network model, and determine the fitness function based on a weighted sum;

[0088] See Figure 3 Step 4: Iteratively update the population, including:

[0089] Update the location of explored and followed individuals to generate an updated population;

[0090] Apply a Gaussian perturbation to individuals in the updated population to generate a perturbed population;

[0091] Calculate the dynamic boundary, update the reverse solutions of individuals in the population, and generate a reverse solution population;

[0092] After merging populations, the optimal set of resistance-related parameters is selected based on the selection probability until the iteration threshold is reached. The set of resistance-related parameters corresponding to the individual with the highest fitness is then used as the resistance control strategy.

[0093] A set of resistance-related parameters corresponds to a set of resistance-related parameters. The temperature of the moving part medium is the temperature of the medium oil (e.g., medium oil) added to the moving object. The working medium delivery pressure is the pressure of the medium oil delivered by the friction-reducing system in the moving object. The working medium flow rate is the amount of medium oil passing through the friction-reducing system in the moving object per unit time. The method for constructing the set of resistance-related parameters is as follows: obtaining the range of resistance-related parameters, which includes the range of moving part medium temperature, the range of working medium delivery pressure, and the range of working medium flow rate; the range of resistance-related parameters is obtained by those skilled in the art based on the technical parameters of the moving object; a value is randomly selected from each range of the resistance-related parameters to construct a set of resistance-related parameters, and a total of n sets of resistance-related parameters are constructed, all of which are different.

[0094] Methods for setting population size include:

[0095] Population size ranges are defined based on historical test data; the population size intervals are dynamically adjusted by the segmentation ratio, and fitness comparison results are calculated; the size intervals are then shrunk based on the fitness comparison results until the interval width is less than a threshold, and the average of the maximum and minimum values ​​in the updated population size range is taken as the population size; the details are as follows:

[0096] Step 201: Collect c sets of historical test data. The historical test data are the test data obtained at historical moments, where c is an integer greater than 1.

[0097] Step 202: Set different population sizes for all historical test data of group c, with the population size ranging from [1, c]; set the split ratio L and the number of iterations L'. In this embodiment, L = 0.618 is preferred, and the number of iterations L' can be set by those skilled in the art according to the actual situation.

[0098] Step 203: Based on the partitioning ratio h, divide the population size into two new points ζ1 and ζ2, ζ1 = 1 + (1 - L)(c - 1), ζ2 = 1 + L(c - 1);

[0099] Step 204: Update the population size according to δ1 and δ2 respectively, perform iteration according to the updated population size, and obtain the fitness corresponding to the position of the individual with the highest fitness; mark the fitness obtained according to δ1 as the first fitness, and mark the fitness obtained according to δ2 as the second fitness;

[0100] Step 205: If the first fitness is greater than or equal to the second fitness, the minimum value of the updated population size range is the minimum value of the population size range before the update, and the maximum value is the point corresponding to the second fitness; if the first fitness is less than the second fitness, the minimum value of the updated population size range is the point corresponding to the first fitness, and the maximum value is the point corresponding to the second fitness and the maximum value of the population size range before the update; for example, after the population size range [1,c] is updated through points δ1 and δ2, if the first fitness is greater than or equal to the second fitness, the updated population size range is [1,δ2], and if the first fitness is less than the second fitness, the updated population size range is [δ1,c].

[0101] Step 206: Preset a width threshold. The width threshold is preset by those skilled in the art based on the actual situation. Subtract the minimum value from the maximum value in the updated population size range to obtain the interval width. Compare the interval width with the width threshold. If the interval width is greater than or equal to the width threshold, the updated population size range is re-divided into two new points δ1 and δ2 according to the segmentation ratio h, and the process returns to step 204. If the interval width is less than the width threshold, the average of the maximum and minimum values ​​in the updated population size range is taken as the population size Z.

[0102] The first individual in the population is the exploratory individual, and the remaining individuals are the follower individuals; the range of parameter labels and the range of the one-dimensional search space are both [1, n].

[0103] The fitness function is expressed as: f = ω1 × sz + ω2 × sx; where f is fitness, sz is screening quality, sx is screening efficiency, and ω1 and ω2 are preset weight coefficients; screening quality refers to the material quality control capability in the process of classifying and screening moving objects, and screening efficiency refers to the quantity of material passing through the sieve per unit time in the process of classifying and screening moving objects; the specific values ​​of the weight coefficients can be set according to the actual situation. The weight coefficients reflect the degree of influence of screening quality and screening efficiency. Those skilled in the art can preset the corresponding weight coefficients according to the actual degree of influence of screening quality and screening efficiency in order to accurately evaluate the comprehensive screening performance of moving objects.

[0104] The screening quality and screening efficiency are obtained as follows: the test data and the set of resistance-related parameters corresponding to the parameter labels of individual positions are used as analysis data. The analysis data are then input into the trained quality prediction model and efficiency prediction model to predict the corresponding screening quality and screening efficiency. The specific training process of the quality prediction model and efficiency prediction model is the same as that of the resistance assessment model, and both are deep belief network models.

[0105] The methods for updating the exploration of individual locations include:

[0106] The exploration individuals, second position, and last position are determined by fitness ranking;

[0107] The position update of the explored individual is based on a randomly weighted offset between the position of the individual with the highest fitness and the second or last position in the population. The offset direction is determined by a random threshold, and the offset is dynamically adjusted by incorporating the random perturbation amplitude of the iteration count to update the position of the explored individual; as follows:

[0108] Sort all individuals in the population by fitness from highest to lowest. The position of the individual with the highest fitness is the next fitness level after the individual with the highest fitness, and the position of the individual with the lowest fitness is the last fitness.

[0109]

[0110] In the formula, X new To explore individual positions, F represents the position of the individual with the highest fitness, g(G) is a random value in a g-distribution with G iterations, W' is the second position, and W” is the last position. A random number in the range [-0.5, 0.5]. It is a random number in [0,1].

[0111] By randomly selecting the second or last position, the algorithm avoids getting trapped in local optima; by combining the randomness of the g(G) distribution with the correlation of the number of iterations, the breadth and depth of the search are balanced; random coefficients Limit the offset range ([﹣0.5,0.5]) to ensure that the position update is within a reasonable range and prevent excessive deviation from the current optimal region.

[0112] Methods for updating the location of following individuals include:

[0113] The new position of a follower is generated by averaging its current position with the position of the previous individual, achieving gradual convergence of positions among individuals within the population. Local cooperation mechanisms balance individual differences, preventing the population from prematurely converging to a single region, while maintaining population stability to support continuous optimization.

[0114] Methods for generating perturbed populations include:

[0115] Gaussian noise is introduced, and Gaussian noise is added to each individual in the updated population to generate new individuals. All new individuals form a perturbed population; the expression for the new individuals is: In the formula, Let X be the position of the i-th new individual. iLet N(0,σ) be the position of the i-th individual, and let N(0,σ) be Gaussian noise with a mean of 0 and a standard deviation of σ, i∈[1,Z]. The standard deviation of the Gaussian noise is set by those skilled in the art according to the required perturbation level. Controllable randomness is introduced to enhance population diversity and help the algorithm escape the local optimum trap. At the same time, the search efficiency is guaranteed by the standard deviation constraint of the perturbation amplitude (such as small-range perturbation).

[0116] Methods for generating a reverse solution population include:

[0117] The dynamic boundary is [θ1, θ2], where θ1 is the parameter label with the smallest value corresponding to the individual position in the population, and θ2 is the parameter label with the largest value corresponding to the individual position in the population, and 1≤θ1<θ2≤n. The dynamic boundary can make up for the deficiency of the fixed boundary in not being able to save search experience, which is beneficial to shorten the search time.

[0118] Methods for generating inverse solutions include:

[0119]

[0120] In the formula, Let be the reverse solution for the i-th individual. It is a random number in [0,1].

[0121] If the generated inverse solution is greater than or less than the dynamic boundary, it is marked as a transcendental solution. A value within the dynamic boundary is randomly generated using a random number function, and the randomly generated value is assigned to the transcendental solution.

[0122] All reverse solutions form a reverse solution population. Dynamic boundary adaptation adjusts the range of solutions in real time according to the current population state to avoid the limitations of fixed boundaries; symmetric solutions are generated using reverse operations to explore potential optimal regions and break through local optima; out-of-bounds solutions are forcibly reset to ensure the rationality of solutions while maintaining search efficiency.

[0123] It should be noted that the reason for adopting the above method to formulate the resistance control strategy is as follows:

[0124] 1. Systematic and flexible: By constructing and dynamically updating a set of resistance-related parameters, different parameter combinations can be explored systematically, improving the flexibility of the strategy;

[0125] 2. Adaptive optimization: The introduction of the fitness function ensures that the algorithm can adjust its strategy based on real-time feedback, thereby optimizing performance;

[0126] 3. Diversity maintenance: By using methods such as Gaussian perturbation and reverse solution, the diversity of the population can be maintained during the optimization process, avoiding getting trapped in local optima;

[0127] 4. Iterative Improvement: By setting the number of iterations and dynamic boundaries, the strategy can be continuously improved to adapt to changing operating conditions;

[0128] 5. Data-driven decision-making: Combining real-time data feedback and fitness assessment makes the final resistance control strategy more scientific and effective.

[0129] The feedback optimization module dynamically adjusts the resistance-related parameters based on the feedback data and re-formulates the resistance control strategy. The feedback data consists of multi-dimensional operating status data after the resistance control is completed.

[0130] Methods for revising resistance control strategies include:

[0131] After resistance regulation is completed, the multidimensional operational state data of the moving object is labeled as feedback data, and the multidimensional operational state data of the moving object collected in real time is labeled as real-time data. The entropy value of the moving object is calculated based on the feedback data, and the presence of resistance anomalies is assessed. If no resistance anomalies are found, the resistance regulation strategy is not reformulated. If resistance anomalies still exist, the set of resistance-related parameters in the resistance regulation strategy and the real-time data are input into the trained parameter prediction model to predict the corresponding prediction data. The parameter prediction model includes a frequency prediction model, an amplitude prediction model, a shaft temperature prediction model, and a screen temperature prediction model. The prediction data includes vibration frequency, vibration amplitude, bearing temperature, and screen temperature. The frequency prediction model is used to predict vibration frequency, the amplitude prediction model is used to predict vibration amplitude, the shaft temperature prediction model is used to predict bearing temperature, and the screen temperature prediction model is used to predict screen temperature. The specific training process for each model in the parameter prediction model is consistent with the specific training process for the resistance assessment model, and both are deep belief network models.

[0132] Subtract the corresponding data from the feedback data from each data point in the forecast data to obtain the unadjusted amount for each data point; subtract the corresponding data from the real-time data from each data point in the forecast data to obtain the amount to be adjusted for each data point; divide the unadjusted amount for each data point in the forecast data by the amount to be adjusted to calculate the unadjusted rate for each data point in the forecast data; add 1 to each unadjusted rate to obtain the adjustment rate for each data point in the forecast data.

[0133] The vibration frequency, vibration amplitude, bearing temperature, and screen temperature in the real-time data are multiplied by the corresponding adjustment rate to obtain the actual data; the screen load in the feedback data is also added to the actual data, and the dynamic change rate index of the moving object is extracted from the actual data and marked as the actual characteristic data; based on the actual data and the actual characteristic data, the resistance control strategy is re-formulated.

[0134] It should be noted that the purpose of revising the resistance control strategy is to accurately identify the control elements that need to be optimized. For example, some parameters may not be well adjusted and need to be emphasized and optimized in the new strategy; to improve the targeting of the control strategy by customizing a more efficient new strategy based on the difference between the existing control effect and the current requirements; and to achieve dynamic optimization of the control strategy by using closed-loop iteration to ensure that the resistance control strategy is consistent with the dynamic performance of the equipment, thereby improving the quality of resistance control.

[0135] This embodiment comprehensively monitors the operating status of a moving object through multi-parameter acquisition, uses entropy analysis to initially assess the resistance status, extracts feature data and combines it with deep learning technology to further evaluate the resistance status and promptly identify abnormal resistance conditions. It employs a swarm intelligence optimization algorithm to formulate a precise resistance control strategy and performs dynamic feedback adjustment. This effectively integrates and utilizes multiple collected data points and performs accurate calculations of resistance-related parameters, achieving quantitative and precise friction reduction for the moving object, improving its operating efficiency and service life, reducing the risk of failure, and providing technical support for increasing production capacity and ensuring safety.

[0136] Example 2

[0137] Please see Figure 4 As shown, parts not described in detail in this embodiment are described in Embodiment 1. A dynamic data intelligent control method is provided, the method including:

[0138] Real-time acquisition of multi-dimensional operational status data of moving objects;

[0139] Collect historical multidimensional operational status data and calculate the entropy value of moving objects;

[0140] Based on the entropy value of the moving object, a preliminary assessment can be made as to whether the moving object exhibits abnormal resistance.

[0141] If there is an initial resistance anomaly, extract the dynamic rate of change index of the moving object;

[0142] Based on the multidimensional operational status data of the moving object, the dynamic rate of change index, and the preset resistance assessment model, the assessment results are output, including whether the resistance is normal or abnormal.

[0143] If the resistance is determined to be abnormal, a resistance control strategy is generated based on the multi-dimensional operational status data of the moving object, and then executed and data is fed back.

[0144] The resistance-related parameters are dynamically adjusted based on the feedback data, and the resistance control strategy is redefined. The feedback data consists of multi-dimensional operating status data after the resistance control is completed.

[0145] The parameters in the above formulas are all dimensionless.

[0146] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0147] In conclusion, 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 spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A dynamic data intelligent control method, characterized in that, The method comprises the following steps: real-time acquisition of multi-dimensional running state data of a moving object; collection of historical multi-dimensional running state data and calculation of entropy value of the moving object; preliminary assessment of whether the moving object has abnormal resistance based on the entropy value of the moving object; if there is a preliminary resistance abnormality, extracting a dynamic change rate index of the moving object; outputting an evaluation result based on the multi-dimensional running state data of the moving object, the dynamic change rate index and a preset resistance evaluation model, the evaluation result including normal resistance or abnormal resistance; if the resistance is determined to be abnormal, generating a resistance control strategy through a swarm intelligence optimization algorithm based on the multi-dimensional running state data of the moving object, executing and feeding back data; dynamically adjusting resistance correlation parameters according to the feedback data, re-formulating the resistance control strategy, and feeding back data after the resistance control is completed.

2. The dynamic data intelligent control method of claim 1, wherein, The method for generating the resistance control strategy through the swarm intelligence optimization algorithm comprises the following steps: Step 1: Constructing n resistance correlation parameter sets, n being an integer greater than 1, each set containing a random combination of medium temperature of the moving component, working medium delivery pressure and working medium flow, and setting different parameter tags for each resistance correlation parameter set; Step 2: Initializing a population, defining the correspondence between individual positions in a one-dimensional search space and parameter tags, and setting the initial iteration number to 0; Step 3: Calculating an adaptability function, predicting the screening quality and screening efficiency through a deep belief network model, and determining the adaptability function based on a weighted sum; Step 4: Iteratively updating the population, including: updating the positions of exploring individuals and following individuals to generate an updated population; applying Gaussian disturbance to the individuals in the updated population to generate a disturbed population; calculating dynamic boundaries to generate a reverse solution population by reversing the solutions of the individuals in the updated population; merging the populations and screening the optimal resistance correlation parameter set based on a selection probability until an iteration threshold is reached, and taking the resistance correlation parameter set corresponding to the individual with the maximum adaptability as the resistance control strategy.

3. The dynamic data intelligent control method of claim 2, wherein, The method for obtaining the screening quality and the screening efficiency comprises the following steps: taking the test data and the resistance correlation parameter set corresponding to the individual position as analysis data, the test data including the multi-dimensional running state data and the dynamic change rate index of the moving object; and predicting the corresponding screening quality and screening efficiency based on the analysis data and preset quality prediction model and efficiency prediction model, the quality prediction model and the efficiency prediction model both being deep belief network models.

4. The dynamic data intelligent control method of claim 2, wherein, The first individual in the population is an exploring individual, and the remaining individuals are following individuals; the range of the parameter tags and the range of the one-dimensional search space are both [1, n].

5. The dynamic data intelligent control method of claim 2, wherein, The method for setting the population size comprises the following steps: dividing the population size range based on historical test data; dynamically adjusting the population size interval through a segmentation ratio and calculating adaptability comparison results; shrinking the size interval according to the adaptability comparison results until the interval width is less than a threshold value, and taking the mean value of the maximum and minimum values in the updated population size range as the population size.

6. The dynamic data intelligent control method of claim 2, wherein, The method for updating the position of the exploring individual comprises the following steps: determining the exploring individual, the second position and the last position according to the adaptability order; The position of the exploration individual is updated based on a random weighted offset of the position of the individual with the highest fitness and a second position or a last position in the population, a random threshold value is used to determine the offset direction, and a random disturbance amplitude of the number of iterations is combined to dynamically adjust the offset and update the position of the exploration individual; The position of the following individual is updated by averaging the current position of the individual and the position of the previous individual.

7. The dynamic data intelligent control method of claim 6, wherein, The method for calculating the entropy value of the moving object comprises: The same data in the historical multi-dimensional running state data and the multi-dimensional running state data is added to an analysis set, i.e., the data in the analysis set multi-dimensional running state data is one-to-one correspondence; a clustering algorithm is used to cluster the data in each analysis set to obtain each cluster corresponding to each analysis set; the number of data in each cluster is obtained, each data number corresponding to each analysis set is sequentially added to obtain the total number of data of each analysis set; the number of data of each cluster is divided by the total number of data of the corresponding analysis set to obtain the interval probability of each cluster; the set entropy value of each analysis set is calculated according to the interval probability of each cluster in each analysis set; and the set entropy value of each analysis set is taken as the entropy value of the moving object.

8. The dynamic data intelligent control method of claim 7, wherein, The method for preliminarily evaluating whether the moving object has resistance abnormality comprises: The historical entropy value is obtained, the historical entropy value being the entropy value of the moving object calculated at a historical moment; the real-time calculated entropy value of the moving object is marked as a real-time entropy value; each set entropy value in the real-time entropy value is added to the corresponding set entropy value in the historical entropy value to obtain the total entropy value of the moving object corresponding to each data in the multi-dimensional running state data; the number of entropy values in the historical entropy value is counted and one is added to obtain the number of entropy values; each vibration total entropy value is divided by the number of entropy values to obtain the mean value of the entropy value corresponding to each data in the multi-dimensional running state data; each set entropy value in the real-time entropy value and the historical entropy value is subtracted by the corresponding mean value of the entropy value and then squared to obtain an entropy difference value; the entropy difference value corresponding to each data in the multi-dimensional running state data is sequentially added and then divided by the number of entropy values to obtain the variance corresponding to each data in the multi-dimensional running state data; each variance is square rooted to obtain the entropy standard deviation corresponding to each data in the multi-dimensional running state data; the mean value of the entropy value of each data in the multi-dimensional running state data is added by three times the corresponding entropy standard deviation to obtain the entropy threshold value; Each set entropy value in the real-time entropy value is compared with the corresponding entropy threshold value; If each set entropy value is less than or equal to the corresponding entropy threshold value, it is preliminarily evaluated that the moving object does not have resistance abnormality; If there is a set entropy value greater than the corresponding entropy threshold value, it is preliminarily evaluated that the moving object has resistance abnormality.

9. The dynamic data intelligent control method of claim 8, wherein, The method for formulating a resistance control strategy comprises: The real-time collected multi-dimensional running state data is marked as real-time data; The feedback data is used to calculate the entropy value of the moving object, and whether there is resistance anomaly of the moving object is evaluated; if there is no resistance anomaly, the resistance regulation strategy is not re-established; if there is still resistance anomaly, the resistance related parameter set in the resistance regulation strategy and the real-time data are input into the trained parameter prediction model to predict corresponding prediction data; the prediction data includes vibration frequency, vibration amplitude, bearing temperature and screen temperature; wherein the frequency prediction model is used to predict the vibration frequency, the amplitude prediction model is used to predict the vibration amplitude, the shaft temperature prediction model is used to predict the bearing temperature, and the screen temperature prediction model is used to predict the screen temperature; the parameter prediction model includes the frequency prediction model, the amplitude prediction model, the shaft temperature prediction model and the screen temperature prediction model; each model in the parameter prediction model is a deep belief network model; each data in the prediction data is subtracted by the corresponding data in the feedback data to obtain the unadjusted amount of each data; each data in the prediction data is subtracted by the corresponding data in the real-time data to obtain the to-be-adjusted amount of each data; the unadjusted rate of each data in the prediction data is calculated by dividing the unadjusted amount of each data in the prediction data by the to-be-adjusted amount; each unadjusted rate is added by 1 as the adjustment rate of each data in the prediction data; The vibration frequency, vibration amplitude, bearing temperature and screen temperature in the real-time data are multiplied by the corresponding adjustment rate to obtain actual data; the screen load in the feedback data is also added to the actual data, and the dynamic change rate index is extracted from the actual data and marked as actual feature data; the resistance regulation strategy is re-established according to the actual data and the actual feature data.

10. A dynamic data intelligent control system implementing a dynamic data intelligent control method according to any one of claims 1-9, characterized by, Comprise: A data acquisition module for acquiring multi-dimensional running state data of a moving object in real time; A data analysis module for collecting historical multi-dimensional running state data and calculating the entropy value of the moving object; A preliminary evaluation module for preliminarily evaluating whether the moving object has resistance anomaly according to the entropy value of the moving object; A feature extraction module for extracting the dynamic change rate index of the moving object if there is preliminary resistance anomaly; An anomaly evaluation module for outputting an evaluation result based on the multi-dimensional running state data of the moving object, the dynamic change rate index and a preset resistance evaluation model, wherein the evaluation result includes normal resistance or resistance anomaly; A resistance regulation module for generating a resistance regulation strategy by a group intelligence optimization algorithm according to the multi-dimensional running state data of the moving object if the resistance anomaly is identified, and executing and feeding back data; A feedback optimization module for dynamically adjusting resistance related parameters according to the feedback data, re-establishing the resistance regulation strategy, and the feedback data being the multi-dimensional running state data fed back after the resistance regulation is completed.

Images