An intelligent optimization and real-time decision-making method for industrial big data

Abstract

The present invention relates to the field of industrial intelligent decision-making technology, specifically to an intelligent optimization and real-time decision-making method for industrial big data. The platform includes: a node data time distribution analysis module that obtains the operating status parameters of product production nodes from industrial big data and converts them into time series, extracts the characteristic values ​​of the operating status parameters in each time period of the time series, and counts the frequency of occurrence of the characteristic values ​​in the corresponding time period to generate node time period distribution information. In the present invention, by obtaining the time series distribution characteristics of the operating status parameters of each node in industrial production, the change characteristics and frequency distribution of the parameters in different time periods can be accurately reflected. By timely capturing the trend changes of the operating status parameters and the key moments that exceed the preset change threshold, the sensitivity and accuracy of abnormal state detection are improved. By analyzing the upper and lower limits of the range of abnormal fluctuation points and the cross-stage correlation strength, the global coordination of parameter adjustment is optimized.

Description

Technical Field

[0001] The present invention relates to the field of industrial intelligent decision-making technology, and in particular to an intelligent optimization and real-time decision-making method for industrial big data. Background Art

[0002] Intelligent optimization and real-time decision-making methods for industrial big data refer to platforms that conduct real-time multi-source data collection, dynamic optimization, and decision-making for the massive amounts of data generated during industrial production. Combined with the different needs of the industrial production process, dynamic management of the production process and real-time decision support are achieved through an integrated platform.

[0003] When analyzing industrial production data, existing technologies can only optimize based on static parameters or data from a single stage, making it difficult to fully capture the dynamic changes in node operating status parameters. For example, existing methods can usually only analyze operating status parameters based on statistical values ​​or historical data, but cannot accurately extract key change points from time series, resulting in a lag in the response of anomaly detection to key changes. In terms of evaluating the range of parameter fluctuations, existing technologies can only make abnormal judgments and find it difficult to accurately define abnormal states. For example, when a parameter experiences multiple small fluctuations during the production process, existing technologies find it difficult to distinguish whether the fluctuations are within the normal operating range, which may lead to false positives or false negatives. In terms of cross-stage correlation optimization, the analysis of existing technologies is mostly limited to the parameter optimization of a single node or a single stage, making it difficult to coordinate the dynamic adjustment of global production. For example, the adjustment of operating status parameters between different nodes may cause an imbalance in production efficiency between stages due to the lack of global optimization guidance, which can easily lead to limited overall optimization effects and affect the improvement of production efficiency and quality. Summary of the Invention

[0004] The purpose of this invention is to solve the shortcomings of the existing technology and propose an intelligent optimization and real-time decision-making method for industrial big data.

[0005] To achieve the above objectives, the present invention adopts the following technical solutions: An intelligent optimization and real-time decision-making method for industrial big data includes:

[0006] The node data time distribution analysis module obtains the operating status parameters of product production nodes from industrial big data and converts them into time series. It extracts the characteristic values ​​of the operating status parameters in each time period of the time series, counts the frequency of occurrence of the characteristic values ​​in the corresponding time period, and generates node time period distribution information.

[0007] The node distribution dynamic evolution module performs dynamic trend evolution on the time series corresponding to the operating status parameters based on the node time period distribution information, generates dynamic trend evolution results, marks the time points in the dynamic trend evolution results when the operating status parameters change by more than a preset change threshold, and generates node distribution change point information;

[0008] The node fluctuation range assessment module analyzes the time series corresponding to the operating status parameters adjacent to the marked time points in the node distribution change point information, determines the upper and lower limits of the fluctuation range of the target operating status parameters based on the analysis results, marks abnormal change points based on the upper and lower limits of the fluctuation range, and generates node abnormal fluctuation information;

[0009] The node multi-stage association optimization module obtains the abnormal operation state parameter optimization interval from the upper and lower limits of the fluctuation range of the target operation state parameter corresponding to the abnormal change point in the abnormal fluctuation information of the node and the optimal operation range of the preset operation state parameter, analyzes the correlation strength between the abnormal operation state parameter optimization interval and the entire product production node, and generates a cross-stage parameter optimization analysis result;

[0010] The node parameter suggestion generation module determines the adjustment order of different operating status parameters according to the correlation strength of all operating status parameters in the cross-stage parameter optimization analysis results, and generates a node parameter optimization decision report according to the adjustment order.

[0011] As a further solution of the present invention, the step of obtaining the node time period distribution information is specifically as follows:

[0012] Obtain the operating status parameters of product production nodes from industrial big data and convert them into time series. Divide the time series into different time periods and obtain the characteristic values ​​of the operating status parameters in each time period. The characteristic values ​​include maximum value, minimum value and quantile value, and generate the characteristic value set of the operating status parameters.

[0013] The frequency of occurrence of each characteristic value in the characteristic value set of the operating status parameter in the corresponding time period is counted, and a histogram is constructed according to the frequency of occurrence to display the frequency distribution of the operating status parameters. The frequency distribution is smoothed and fitted using the kernel density estimation method to generate the node time period distribution information.

[0014] As a further solution of the present invention, the steps for obtaining the dynamic trend evolution result are specifically as follows:

[0015] Based on the node time period distribution information, the formula is adopted:

[0016]

[0017] In the time series corresponding to the calculation of the operating state parameters, the operating state parameters at the previous moment are S t-1 When the current operating state parameter is S t The conditional probability P(S t |S t-1 ), generate the state change probability of the operating state parameters;

[0018] Among them, A(St-1 ,S t ) is obtained from the statistics of the number of changes of the operating state parameters in the time series corresponding to the operating state parameters, and is the operating state parameter S at the previous moment. t-1 Transfer to the current operating state parameter S t The probability of n is the total number of all operating state parameters, B(S t ,O t ) is obtained based on the frequency of occurrence of each characteristic value in the operating state parameter characteristic value set in the corresponding time period in the current operating state parameter S t The specified operating state parameter O is observed t The observation probability of k represents the index of all operating state parameters;

[0019] According to the state change probability of the operating state parameters, the change rules of the operating state parameters in different periods are analyzed to generate dynamic trend evolution results.

[0020] As a further solution of the present invention, the step of obtaining the node distribution change point information is specifically as follows:

[0021] Based on the dynamic trend evolution result, by calculating the change amount of the operating status parameter in the time series corresponding to the operating status parameter point by point, marking the time point when the change amount exceeds the preset change threshold, wherein the marked time point represents the moment when a key change occurs in the time series corresponding to the operating status parameter, and generating a marked time point set;

[0022] The change types and distribution density corresponding to the operating status parameters in the marked time point set are classified and counted, and the distribution law of each marked time point is analyzed in combination with the time interval of the time point to generate node distribution change point information.

[0023] As a further solution of the present invention, the step of determining the upper and lower limits of the fluctuation range of the target operating state parameter is specifically:

[0024] The time series corresponding to the running state parameters adjacent to the marked time points in the node distribution change point information are analyzed using the formula:

[0025]

[0026] Calculate the speed v at which the target operating state parameter deviates from the current value x to the corresponding mean μ and the fluctuation amplitude σ of the target operating state parameter respectively;

[0027] Among them, k' represents the intensity of the target operating state parameter returning to the mean after deviating from the corresponding mean, is the proportional coefficient, x is the current value of the target operating state parameter, μ represents the mean corresponding to the target operating state parameter, is the instantaneous change rate of the adjacent operating state parameters, β is the instantaneous change rate weight coefficient set according to the instantaneous change rate, D represents the basic value of the fluctuation range of the target operating state parameter, Δx is the fluctuation increment, which represents the difference between the maximum and minimum values ​​in the time series corresponding to the operating state parameter, and γ is the fluctuation range adjustment coefficient set according to the fluctuation increment;

[0028] According to the speed v at which the target operating state parameter deviates from the current value x to the corresponding mean μ and the fluctuation amplitude σ of the target operating state parameter, the formula is adopted:

[0029] [x min ,x max ]=[μ-(σ+v·Δt),μ+(σ+v·Δt)]

[0030] Determine the upper limit x of the fluctuation range of the target operating state parameter max and the lower limit x min ;

[0031] Wherein, Δt represents the difference between adjacent time points in the time series corresponding to the operating state parameters.

[0032] As a further solution of the present invention, the step of obtaining the abnormal node fluctuation information is specifically as follows:

[0033] Comparing the upper and lower limits of the fluctuation range of the target operating state parameter with the operating state parameter adjacent to the marked time point, marking abnormal change points exceeding the upper and lower limits of the fluctuation range according to the comparison results, and generating abnormal change point marking results;

[0034] Based on the abnormal change point marking results, the distribution characteristics of all abnormal change points are classified and sorted to generate node abnormal fluctuation information.

[0035] As a further solution of the present invention, the steps for obtaining the node parameter optimization decision report are specifically as follows:

[0036] Determining the importance of different operating status parameters based on the correlation strength of all operating status parameters in the cross-stage parameter optimization analysis results, allocating operating status parameter adjustment order priorities based on the importance, and generating operating status parameter priority adjustment results;

[0037] The operation status parameter priority adjustment result is displayed in a visual manner to generate a node parameter optimization decision report.

[0038] Compared with the prior art, the advantages and positive effects of the present invention are:

[0039] In the present invention, by obtaining the time series distribution characteristics of the operating status parameters of each node in industrial production, the change characteristics and frequency distribution of the parameters in different time periods can be accurately reflected. By timely capturing the trend changes of the operating status parameters and the critical moments when the preset change threshold is exceeded, the sensitivity and accuracy of abnormal state detection are improved. By analyzing the upper and lower limits of the range of abnormal fluctuation points and the cross-stage correlation strength, the global coordination of parameter adjustment is optimized, making the optimization of operating status parameters at different stages more efficient. At the same time, by adjusting the priority, it can be ensured that the parameter adjustment process has a clear execution order, and the optimization decision is more accurate and efficient. In the multi-stage optimization analysis, the correlation between different operating status parameters is systematically considered, which further improves the effect of global optimization and enhances the adaptability of the industrial production process to complex dynamic changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a platform flow chart of the present invention;

[0041] Figure 2 This is a flow chart of the present invention for obtaining node time period distribution information;

[0042] Figure 3 A flow chart of obtaining dynamic trend evolution results for the present invention;

[0043] Figure 4 This is a flow chart of the present invention for obtaining node distribution change point information;

[0044] Figure 5 A flow chart for determining the upper and lower limits of the fluctuation range of the target operating state parameter of the present invention;

[0045] Figure 6 This is a flow chart of the present invention for obtaining abnormal node fluctuation information;

[0046] Figure 7 A flow chart for obtaining cross-stage parameter optimization analysis results for the present invention;

[0047] Figure 8 This is a flowchart of the present invention for obtaining a node parameter optimization decision report. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0049] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0050] See also Figure 1 , an intelligent optimization and real-time decision-making method for industrial big data includes:

[0051] The node data time distribution analysis module obtains the operating status parameters of product production nodes from industrial big data and converts them into time series. It extracts the characteristic values ​​of the operating status parameters in each time period of the time series, counts the frequency of occurrence of the characteristic values ​​in the corresponding time period, and generates node time period distribution information.

[0052] The node distribution dynamic evolution module performs dynamic trend evolution on the time series corresponding to the operating status parameters based on the node time period distribution information, generates dynamic trend evolution results, marks the time points in the dynamic trend evolution results when the operating status parameters change beyond the preset change threshold, and generates node distribution change point information;

[0053] The node fluctuation range assessment module analyzes the time series corresponding to the operating status parameters adjacent to the marked time points in the node distribution change point information, determines the upper and lower limits of the fluctuation range of the target operating status parameters based on the analysis results, marks abnormal change points based on the upper and lower limits of the fluctuation range, and generates node abnormal fluctuation information;

[0054] The node multi-stage association optimization module obtains the abnormal operation state parameter optimization interval from the upper and lower limits of the fluctuation range of the target operation state parameter corresponding to the abnormal change point in the node abnormal fluctuation information and the optimal operation range of the preset operation state parameter, analyzes the correlation strength between the abnormal operation state parameter optimization interval and the entire product production node, and generates cross-stage parameter optimization analysis results;

[0055] The node parameter recommendation generation module determines the adjustment order of different operating status parameters based on the correlation strength of all operating status parameters in the cross-stage parameter optimization analysis results, and generates a node parameter optimization decision report based on the adjustment order.

[0056] See also Figure 2 ,The specific steps for obtaining node time period distribution information are:

[0057] Obtain the operating status parameters of product production nodes from industrial big data and convert them into time series. Divide the time series into different time periods and obtain the characteristic values ​​of the operating status parameters in each time period. The characteristic values ​​include maximum value, minimum value and quantile value, and generate the characteristic value set of the operating status parameters.

[0058] Extract the operating status parameters of product production nodes from industrial big data, such as the temperature (unit: °C), pressure (unit: MPa) and flow (unit: m 3 / h), etc., by sorting the timestamps of these parameters to form time series data, and dividing the time series into different time periods (such as one period per hour), and calculating the maximum, minimum and quantile values ​​(such as 25%, 50%, and 75% quantiles) of the equipment operation status parameters in each period.

[0059] The frequency of occurrence of each eigenvalue in the set of eigenvalues ​​of the operating status parameters within the corresponding time period is counted, and a histogram is constructed based on the frequency of occurrence to display the frequency distribution of the operating status parameters. The frequency distribution is smoothed and fitted using the kernel density estimation method to generate the node time period distribution information;

[0060] Perform frequency statistics on the acquired eigenvalue set. Taking the device temperature as an example, the maximum value, minimum value and quantile value are grouped and counted. By setting the numerical interval (such as 40℃ to 50℃, 50℃ to 60℃, 60℃ to 70℃, etc.), the frequency of occurrence of the eigenvalues ​​in each interval is counted. Assuming that the statistical result is that the interval 40℃ to 50℃ appears 10 times, the frequency of occurrence 50℃ to 60℃ appears 20 times, the frequency of occurrence 60℃ to 70℃ appears 15 times, and the frequency of occurrence 70℃ to 80℃ appears 5 times, the statistical results are plotted. Histogram, the horizontal axis of the histogram is the temperature range, and the vertical axis is the frequency of occurrence of the corresponding range. The unfitted histogram presents a discrete columnar distribution, reflecting the frequency changes of the parameters in different intervals; the kernel density estimation method is used to smoothly fit the histogram data, for example, the Gaussian kernel function is selected to convert the columnar distribution into a continuous and smooth distribution curve. The fitted distribution curve eliminates the abruptness of the columnar distribution and can more clearly display the frequency change trend of each temperature range, generating node time period distribution information including the time axis and the temperature distribution fitting curve.

[0061] See also Figure 3 ,The specific steps for obtaining the dynamic trend evolution results are:

[0062] Based on the node time period distribution information, the formula is used:

[0063]

[0064] In the time series corresponding to the calculation of the operating state parameters, the operating state parameters at the previous moment are St-1 When the current operating state parameter is S t The conditional probability P(S t |S t-1 ), generate the state change probability of the operating state parameters;

[0065] Among them, A(S t-1 ,S t ) is obtained from the statistics of the number of changes of the operating state parameters in the time series corresponding to the operating state parameters, and is the operating state parameter S at the previous moment. t-1 Transfer to the current operating state parameter S t The probability is calculated by the following formula: N(S t-1 ,S t ) represents the operating status parameter S at the previous moment t-1 Transfer to the current operating state parameter S t The number of changes, is the operating state parameter S at the previous moment t-1 The total number of changes to all operating state parameters, n represents the total number of all operating state parameters (for example, when the operating state parameters are divided into 3 intervals, n = 3), j represents the index of each target operating state parameter (for example, j = 1, 2, 3 corresponds to operating state parameters 1, 2, 3 respectively), the calculation of the total number includes the following steps: t-1 The number of times N(S) the parameters are transferred to each operating state t-1 ,S j ), B(S t ,O t ) is obtained based on the frequency of occurrence of each characteristic value in the operating state parameter characteristic value set in the corresponding time period in the current operating state parameter S t The specified operating state parameter O is observed t The observation probability, for the specified operating state parameter O t For example, if the temperature at a certain moment is observed to be 65°C, then O t =65℃, B(S t ,O t ) is calculated using the following formula: F(O t ,S t ) indicates the current operating state parameter S t The specified operating state parameter O is observed t frequency, Indicates the current operating state parameter S t The total frequency of all operating state parameters observed under i Indicates the current operating state parameter S tThe i-th observation value observed under the current running state parameter, m represents the total number of observation values ​​in the current running state parameter, t represents the time point of each running state parameter in the time series corresponding to the running state parameter, Indicates all operating status parameters S k The sum of is used to normalize the probability of state change, and the operating state parameter set S k It is determined by dividing the operating status parameter intervals generated from the operating status parameter characteristic value set (for example, the temperature interval 50℃-60℃ is operating status parameter 1, 60℃-70℃ is operating status parameter 2, and 70℃-80℃ is operating status parameter 3), and k represents the index of all operating status parameters.

[0066] Assume that the operating state parameter 1 at the previous moment (temperature range 50℃-60℃) is changed to the operating state parameter 2 at the current moment (temperature range 60℃-70℃), the observed value O t =65℃.

[0067] Operation state parameter transition probability A(S t-1 ,S t ) It is necessary to count the number of times the running state parameter 1 is transferred to the running state parameter 2, as well as the total number of times the running state parameter 1 is transferred to all the running state parameters: if the number of times from the running state parameter 1 to the running state parameter 2 is N(S t-1 ,S t )=30;If the number of times from the running state parameter 1 to the running state parameter 3 is N(S t-1 , S3)=10;If the number of times the running state parameter 1 is maintained at the running state parameter 1 is N(S t-1 ,S1)=60. The total number of times is:

[0068] but:

[0069] Observation probability B(S t ,O t ) indicates that the parameter S is in the running state t The parameter value O is observed t The probability of . Use the formula to calculate: If the frequency of the parameter value 65°C is observed under the operating state parameter 2, the frequency of the parameter value 65°C is F(O t =65,S t =2)=6; if the frequency of the parameter value 62°C is observed under the operating state parameter 2, F(O i =62,S t =2)=4; if the total frequency under operating state parameter 2 is Substituting into the formula:

[0070] Sum all possible operating state parameters: A(S t-1 ,S1)=0.6,B(S1,O t )=0.4 (the probability of observing 65℃ under operating state parameter 1 is 0.4); A(S t-1 ,S2)=0.3,B(S2,O t )=0.6;A(S t-1 ,S3)=0.1,B(S3,O t )=0.2 (the probability of observing 65°C under operating state parameter 3 is 0.2). Calculation:

[0071] Substitute into the formula:

[0072] The results show that the probability of changing from operating state parameter 1 (temperature range 50℃-60℃) to operating state parameter 2 (temperature range 60℃-70℃) is 40.91%. The same is true for the calculation of pressure and flow operating state parameters.

[0073] The hidden Markov model combines the state transition probability A(S t-1 ,S t ) and the observation probability B(S t ,O t ), inferring and probabilistically calculating the operating state parameters at different moments. Specifically, the hidden Markov model regards the operating state parameters as implicit variables, and calculates the conditional probability P(S) of the state change at each moment in the time series. t |S t-1 ) and uses the previous moment's operating state parameters and observations to update the current moment's operating state parameters, generating a probability sequence of operating state parameter changes at each moment. This process combines the dynamic evolution trend of time series with the distribution characteristics of observed values, making the analysis of state changes more consistent with actual operating laws.

[0074] According to the state change probability of the operating state parameters, the changing rules of the operating state parameters in different periods are analyzed to generate dynamic trend evolution results;

[0075] According to the calculated operating state parameter change probability P(S t |S t-1) = 0.4091. Combined with the changing trends of the operating state parameters in the time series, the changing patterns of the operating state parameters are analyzed. Assume that the operating state parameters in the time series change as follows: at t1, the operating state parameter is S1 (temperature range 50°C-60°C); at t2, the operating state parameter is S2 (temperature range 60°C-70°C); and at t3, the operating state parameter is S3 (temperature range 70°C-80°C). From t1 to t2, the operating state parameter changes from S1 to S2, with a calculated probability of change of 40.91%. Combined with time series analysis, this change in the equipment operating state parameters conforms to a continuous temporal evolution trend, with no apparent stagnation of the operating state parameters. Observations from the set of operating state parameter eigenvalues ​​show that the temperature increases from the range at t1 (50°C-60°C) to the range at t2 (60°C-70°C), indicating that the equipment operating state parameters continue to evolve toward a higher temperature range. From t2 to t3, the operating state parameter changes from S2 to S3, with a calculated probability of change of 60%. Combined with time series analysis, this change probability is higher than the probability from t1 to t2, indicating an increase in the rate of change of the operating parameters. Observational statistics show that the temperature increased from the range of 60°C-70°C at t2 to the range of 70°C-80°C at t3, and the equipment's operating parameters did not fluctuate repeatedly during this period. Based on this analysis, the changing pattern of the operating parameters is as follows: the parameters continuously transitioned to higher temperature ranges, without stagnation or decline. The changes in different time periods exhibited a certain degree of continuity, with a probability of 40.91% from t1 to t2 and 60% from t2 to t3, indicating that the evolution rate of the operating parameters gradually increased, while the reverse trend decreased. Throughout the entire time series, the operating parameters showed no significant repetitions, meaning that at each stage of the time series, the operating parameters continuously evolved into new ranges, indicating a relatively stable operating trend. Significant repetitions could indicate that the operating parameters were potentially affected by external disturbances or internal instability.

[0076] See also Figure 4 ,The specific steps for obtaining node distribution change point information are:

[0077] Based on the dynamic trend evolution results, the change amount of the operating status parameter in the time series corresponding to the operating status parameter is calculated point by point, and the time point when the change amount exceeds the preset change threshold is marked. The marked time point represents the moment when a key change occurs in the time series corresponding to the operating status parameter, and a marked time point set is generated;

[0078] To mark time points in the dynamic trend evolution results where the change in operating status parameters exceeds a preset change threshold, first, calculate the change in the operating status parameters point by point based on the dynamic trend evolution results. For example, in a temperature parameter time series: suppose the temperature is 60°C at time t1 and 75°C at time t2. The change is calculated as: 75 - 60 = 15. The preset temperature change threshold range is ±10°C. Here, a change of 15°C exceeds the threshold, so t2 is marked as a significant change time point. Similarly, the change is calculated for each time point in the time series and compared with the preset threshold. For example, the temperature changes from 75°C to 85°C from t2 to t3, a change of 10°C, which does not exceed the threshold. From t3 to t4, the temperature changes from 85°C to 100°C, a change of 15°C, which exceeds the threshold, so t4 is marked. Through this screening, all time points exceeding the threshold are marked. The marking result includes the time point and its corresponding change. For example, the marking results may be: {time point t2, change 15}, {time point t4, change 15}. These marked time points indicate the moments when important changes occurred in the time series corresponding to the operating status parameters and require further analysis.

[0079] Classify and count the change types and distribution density of the operating status parameters in the marked time point set, analyze the distribution pattern of each marked time point based on the time interval of the time point, and generate node distribution change point information;

[0080] Node distribution change point information is generated based on a set of marked time points. First, the marked time points are classified and counted. For example, they are categorized by the change parameter (e.g., temperature) and the change type (e.g., temperature rise): at time point t2, the change type is a temperature rise, and the change amount is 15; at time point t4, the change type is also a temperature rise, and the change amount is 15. Next, the distribution density of the marked time points in the time series is analyzed. This distribution density determination requires calculation based on the specific time unit. Assuming the time unit is "minutes," the time points in the time series record the operating status parameters at minute intervals, for example, t1 is 1 minute, t2 is 2 minutes, and so on. Assume that density is determined based on the following criteria: if the time interval between marked time points does not exceed 5 minutes, and there are three or more marked time points within that time period, then the distribution is considered "highly dense." For example, if the marked points are t2, t3, and t4, and their time intervals are t3-t2 = 2 minutes and t4-t3 = 3 minutes, respectively, and the total time period is within 5 minutes, and contains three marked points, then the distribution is considered "highly dense." If the marked points are t2, t5, and t7, and their time intervals are t5-t2 = 10 minutes and t7-t5 = 15 minutes, respectively, and their time intervals are all greater than 5 minutes, then the distribution is considered "lowly dense." Finally, the marked time points are combined with the density information to generate node distribution change point information. For example: {time point t2, parameter change type "heating up", change amount 15, distribution density "high"}; {time point t4, parameter change type "heating up", change amount 15, distribution density "low"}. Similarly, the markings for pressure parameters and flow parameters will also be divided based on their respective operating ranges. For example, pressure parameters can be divided into low pressure, medium pressure, and high pressure, and flow parameters can be divided into low flow, medium flow, and high flow, so that their change points can be classified, recorded, and marked.

[0081] See also Figure 5 ,The steps for determining the upper and lower limits of the fluctuation range of the target operating state parameters are as follows:

[0082] The time series corresponding to the operating status parameters adjacent to the marked time points in the node distribution change point information are analyzed using the formula:

[0083]

[0084] Calculate the speed v at which the target operating state parameter deviates from the current value x to the corresponding mean μ and the fluctuation amplitude σ of the target operating state parameter respectively;

[0085] Among them, k' represents the intensity of the target operating state parameter returning to the mean after deviating from the corresponding mean. It is a proportional coefficient used to control the proportion of mean recovery in the drift rate. The value range of k' is usually obtained by fitting historical data. For example, under a device operating state parameter, the relationship between the deviation value of the parameter (x-μ) and its regression speed v is recorded. The actual value of k' can be determined by fitting the curve using the least squares method. In order to record the relationship between the deviation value of the operating state parameter and its regression speed, real-time data of the operating state parameter at different times can be collected. First, the deviation value of the operating state parameter from the target mean at each time point is calculated, such as the difference between the recorded temperature value and the target mean. At the same time, its regression speed is calculated between adjacent time points, that is, the rate of change per unit time is obtained by differencing the time series data. These data points are organized into a set of data pairs, each pair of data includes the deviation value and the corresponding regression speed. Using the least squares method to fit the curve, the best fitting straight line for these data points can be obtained. The slope of the fitting curve is the actual value of the mean recovery intensity k'. For example, if within a period of time The temperature deviation from the target mean is recorded as it gradually decreases from 5°C to 1°C, and the corresponding regression rates are recorded as 1.25°C / minute and 0.25°C / minute, respectively. By fitting these data points, the linear relationship between the deviation and the regression rate can be intuitively determined, thereby calculating k'. In addition, k' can be adjusted based on the sensitivity of the device's operation. For example, a larger k' value is set for devices sensitive to temperature fluctuations to ensure rapid regression to the mean, while a smaller k' value can be used for devices with stable operation. For sensitive devices (such as those in environments with high temperature control precision), k' is typically in the range of 0.3 to 0.5. For non-sensitive devices (such as those with low-frequency fluctuations in environmental parameters), k' can range from 0.1 to 0.3. x is the current operating status parameter value, collected and recorded in real time by a sensor. For example, a temperature sensor can directly record the current device operating temperature in °C. μ represents the mean value corresponding to the target operating status parameter, which can be obtained by statistical calculation of a large amount of historical data. For example, in industrial equipment, the target operating temperature range is 60°C-80°C. Long-term operating data can yield a mean of μ = 70°C. It is the instantaneous rate of change of adjacent operating state parameters, indicating the speed of change of operating state parameters per unit time, and is calculated by the following formula: x(t1) and x(t2) are the operating state parameter values ​​at two adjacent time points, and t2-t1 is the time interval between the two time points. The instantaneous rate of change can be obtained by directly calculating the parameter increment and the time increment. The unit is the unit of the operating state parameter per unit time (such as °C / minute). For example, the temperature at time point t1 = 1 minute is 60 °C, and the temperature at time point t2 = 2 minutes is 75 °C, then: β is the instantaneous change rate weight coefficient set according to the instantaneous change rate, which indicates the degree of influence of the instantaneous change on the drift rate. The size of β directly determines the proportion of the instantaneous change rate in the drift rate. The value of β is usually obtained by analyzing the instantaneous change rate (such as ) to the drift speed. For example, by observing the improvement effect of the overall regression speed when the instantaneous change rate is high, a reasonable weight value can be determined. In addition, β can be adjusted in combination with the actual operation characteristics of the equipment. For example, for fast-response equipment, the effect of the instantaneous change rate is more significant, and the value of β will be set higher. For equipment with smooth fluctuations, the value of β is set lower accordingly. For example, the operating temperature (operating status parameter) of industrial welding equipment will fluctuate rapidly with environmental changes, and the instantaneous change rate has a greater impact on the drift rate. In this case, by analyzing the historical time series, it is found that when the instantaneous change rate When the drift rate exceeds 10℃ / min, the fluctuation range of the drift rate increases significantly. According to this situation, the value of β can be set to 0.5~0.7 to ensure that the instantaneous change has a higher weight. In the ambient temperature control equipment, the operating temperature changes slowly, and the instantaneous change has little effect on the drift rate. The historical data shows Most of the time, the temperature remains below 1°C / minute. The drift rate mainly depends on the mean reversion process. Therefore, for such equipment, the value of β can be set to 0.2-0.4. D represents the basic value of the fluctuation amplitude of the target operating state parameter, which is obtained by statistically analyzing the variance of the time series corresponding to the operating state parameter. For example, within a period of time, the temperature of the equipment fluctuates between 60°C and 80°C. The calculation process is: x i' is the current operating status parameter value at the i'th time point, is the mean value of the current operating state parameter value x, n' is the number of time points in the time series corresponding to the operating state parameter, Δx is the fluctuation increment, which represents the difference between the maximum value max(x) and the minimum value min(x) in the time series corresponding to the operating state parameter, and is calculated by the following formula: Δx = max(x) - min(x), γ is the fluctuation range adjustment coefficient set according to the fluctuation increment, which represents the contribution ratio of the fluctuation increment Δx to the fluctuation amplitude. The value of γ is determined by fitting the contribution relationship of the fluctuation increment to the fluctuation amplitude. For example, by analyzing the proportion of the maximum fluctuation increment to the fluctuation amplitude in historical data, a reasonable γ value can be determined. In addition, for parameters with larger fluctuations, the value of γ can be set higher, while for parameters with smaller fluctuations, the value of γ will be lower. For example, in chemical equipment In the example, the operating pressure (operating status parameter) fluctuates widely, potentially rising from 100 kPa to 200 kPa within 10 minutes. This dramatic fluctuation requires the fluctuation increment Δx to occupy a higher weight in the fluctuation amplitude calculation. Therefore, by analyzing historical data, it is found that when Δx>50 kPa, the fluctuation amplitude trend and the increment show a strong linear relationship. Therefore, γ can be set to 0.4-0.5. In temperature control equipment, the operating temperature (operating status parameter) fluctuates relatively small, possibly varying between 60°C and 70°C. The fluctuation increment Δx has little effect on the fluctuation amplitude, which mainly depends on the random fluctuation intensity D. Therefore, by analyzing historical data, it is found that the fluctuation amplitude trend has a weak correlation with Δx. Therefore, γ can be set to 0.2-0.3.

[0086] Assume the temperature time series is as follows: time point t1 = 0 minutes, temperature x(t1) = 60°C; time point t2 = 1 minute, temperature x(t2) = 75°C; time point t3 = 2 minutes, temperature x(t3) = 85°C. Assume the historical operating target mean μ = 70°C; mean reversion strength k' = 0.25; instantaneous change rate weight coefficient β = 0.5; fluctuation range adjustment coefficient γ = 0.3;

[0087] Calculate the basic value of fluctuation range D:

[0088] The basic value of the fluctuation amplitude D≈116.67℃.

[0089] Calculate the fluctuation increment Δx: from time point t1 to t2: Δx = x(t2) - x(t1) = 75 - 60 = 15°C;

[0090] Time point t2 to t3: Δx = x(t3) - x(t2) = 85 - 75 = 10°C;

[0091] Calculate the instantaneous rate of change

[0092] Time point t1 to t2:

[0093] Time point t2 to t3:

[0094] Calculate the speed v at which the temperature operating state parameter deviates from the current value x to the corresponding mean μ and then regresses:

[0095] Time point t1 to t2: Current temperature x = 75°C; target mean μ = 70°C; instantaneous rate of change Substituting into the formula: v = -0.25·(75-70) + 0.5·15 = 6.25°C / min;

[0096] Time point t2 to t3: Current temperature x = 85°C; target mean μ = 70°C; instantaneous rate of change Substituting into the formula: v = -0.25·(85-70)+0.5·10 = 1.25°C / min;

[0097] Calculate the fluctuation amplitude σ:

[0098] Time point t1 to t2: D = 116.67°C, k' = 0.25, γ = 0.3, Δx = 15°C.

[0099] Substituting into the formula:

[0100] Time point t2 to t3: D = 116.67°C, k' = 0.25, γ = 0.3, Δx = 10°C.

[0101] Substituting into the formula:

[0102] The results show that from time point t1 to t2: the fluctuation increment Δx = 15℃; the instantaneous change rate The temperature operating state parameter deviates from the current value x to the corresponding mean μ and then regresses at a speed v = 6.25°C / minute; the fluctuation amplitude σ of the temperature operating state parameter is ≈ 15.42°C. Time point t2 to t3: Fluctuation increment Δx = 10°C; instantaneous rate of change The speed at which the temperature operating state parameter deviates from the current value x to the corresponding mean μ and then regresses is v = 1.25°C / minute; the fluctuation amplitude of the temperature operating state parameter σ≈15.37°C.

[0103] The drift rate and fluctuation amplitude of pressure and flow can also be assessed using calculations similar to those used for temperature. The drift rate indicates how quickly pressure or flow returns to its target mean after deviating from it, while the fluctuation amplitude indicates the range of pressure or flow fluctuation within a specific time interval. In practical applications, time series data for pressure or flow can be used as input. By extracting pressure or flow values ​​at different time points, their instantaneous rate of change and deviation from the target mean are calculated. Similar to the temperature calculation process, the drift rate can be calculated by analyzing the deviation and instantaneous rate of change of pressure or flow from the target mean, while the fluctuation amplitude can be used to reflect its range by using the variance or incremental changes in historical pressure or flow data. Similarly, the pressure fluctuation amplitude can be assessed based on the incremental range of pressure changes and the variance of historical pressure. For flow, the time series may record flow per unit time. By calculating the incremental change of flow within a time interval and combining it with the deviation from the target flow mean, the drift rate can be calculated. The flow fluctuation amplitude can be characterized by the variance or local incremental changes in historical flow data to characterize its dynamic range. Similar to temperature calculation, by using time series-based pressure and flow data, combined with the instantaneous rate of change and the deviation from the target mean, the drift rate and fluctuation amplitude can be calculated respectively, providing a basis for the dynamic change analysis of pressure and flow and the evaluation of operating status parameters.

[0104] According to the speed v of the target operating state parameter after it deviates from the current value x to the corresponding mean μ and the fluctuation amplitude σ of the target operating state parameter, the formula is adopted:

[0105] [x min ,x max ]=[μ-(σ+v·Δt),μ+(σ+v·Δt)]

[0106] Determine the upper limit x of the fluctuation range of the target operating state parameter max and the lower limit x min ;

[0107] Wherein, Δt represents the difference between adjacent time points in the time series corresponding to the operating status parameters. Assuming that time points t1 = 0 minutes and t2 = 1 minute, then: Δt = t2 - t1 = 1 minute;

[0108] The setting parameters were μ=70° C.; σ=15.42° C. and σ=15.38° C.; v=6.25° C. / min and v=1.25° C. / min; Δt=1 min.

[0109] Substitute into the formula: Time point t1 to t2: x min =μ-(σ+v·Δt)=70-(15.42+6.25)=48.33℃; x max=μ+(σ+v·Δt)=70+(15.42+6.25)=91.67℃;

[0110] The results show that the temperature fluctuation range from time point t1 to t2 is [48.33℃, 91.67℃].

[0111] Time point t2 to t3: x min =μ-(σ+v·Δt)=70-(15.38+1.25)=53.37℃; x max =μ+(σ+v·Δt)=70+(15.38+1.25)=86.63℃;

[0112] The results show that the temperature fluctuation range from time point t2 to t3 is [53.37℃, 86.63℃].

[0113] The upper and lower limits of the pressure and flow fluctuation range can be determined in the same manner as temperature. The core method involves dynamic calculation based on the drift rate and fluctuation amplitude. For the operating parameters of pressure or flow, first obtain a target mean value from historical operating data. The drift rate formula is then used to calculate the rate of deviation and regression over different time periods. The fluctuation amplitude of the dynamic range is then calculated using the fluctuation amplitude formula. The target mean value for pressure or flow is then adjusted to a dynamic upper and lower limit interval. The upper and lower limits of this interval are determined based on the fluctuation amplitude and drift rate adjustments to the mean value. For example, for pressure, the drift rate and fluctuation amplitude can be calculated based on pressure changes at adjacent time points during the operating period. These dynamic parameters are then combined with the historical pressure mean to determine the dynamic pressure fluctuation range. Similarly, for flow, the drift rate and fluctuation amplitude are calculated from time series flow data. These data are then combined with the target flow mean to determine the upper and lower limits of the dynamic flow rate. Therefore, the calculation logic for the fluctuation range for both pressure and flow can be fully applied to the temperature process, requiring only parameter adjustments based on the specific pressure or flow data.

[0114] See also Figure 6 ,The specific steps for obtaining node abnormal fluctuation information are:

[0115] Compare the upper and lower limits of the fluctuation range of the target operating status parameter with the operating status parameter adjacent to the marked time point, mark abnormal change points that exceed the upper and lower limits of the fluctuation range according to the comparison results, and generate abnormal change point marking results;

[0116] Suppose the temperature time series adjacent to a certain marked time point is {65°C, 72°C, 90°C, 55°C, 80°C}, with a fluctuation range of [60°C, 85°C]. For the first temperature value, 65°C, check whether it falls within the range of 60°C, 85°C. If it does, mark it as "normal." For the second temperature value, 72°C, check whether it falls within the range of [60°C, 85°C]. If it does, mark it as "normal." For the third temperature value, 90°C, check whether it falls within the range of [60°C, 85°C]. If it exceeds the upper limit of 85°C, mark it as "above the upper limit" and record the excess value as 90-85 = 5°C. For the fourth temperature value, 55°C, check whether it falls within the range of [60°C, 85°C]. If it falls below the lower limit of 60°C, mark it as "below the lower limit" and record the excess value as 60-55 = 5°C. For the fifth temperature value, 80°C, we check to see if it falls within the range of [60°C, 85°C]. If it does, we mark it as "Normal." The results are: Time point 1: 65°C, Normal; Time point 2: 72°C, Normal; Time point 3: 90°C, 5°C above the upper limit; Time point 4: 55°C, 5°C below the lower limit; Time point 5: 80°C, Normal. The same analysis is performed for the operating status parameters of pressure and flow, providing data support for the subsequent generation of abnormal node fluctuation information.

[0117] Based on the abnormal change point marking results, the distribution characteristics of all abnormal change points are classified and sorted to generate node abnormal fluctuation information;

[0118] The marking results are integrated, and the operating status parameters and time point information marked as abnormal change points are extracted. The abnormal points below the lower limit of the fluctuation range and above the upper limit of the fluctuation range are sorted into different categories. The distribution pattern of the abnormal points in the time series is analyzed, and the excess amplitude of each abnormal change point is recorded. At the same time, the complete node abnormal fluctuation information is generated in combination with the time point information. The node abnormal fluctuation information includes the abnormal type, time point, abnormal value of the operating status parameter and the corresponding deviation, which is used for subsequent analysis of the change trend of the node operating status parameters and the law of abnormal occurrence.

[0119] See also Figure 7 , the specific steps for obtaining the cross-stage parameter optimization analysis results are:

[0120] Compare the upper and lower limits of the fluctuation range of the operating state parameters corresponding to the abnormal change points in the node abnormal fluctuation information with the optimal operating range of the preset operating state parameters, and obtain the optimized range of the abnormal operating state parameters from the comparison results;

[0121] The upper and lower limits of the fluctuation range of the operating state parameters corresponding to the abnormal change point in the node abnormal fluctuation information are compared with the preset optimal operating range of the operating state parameters. The operating state parameters and their upper and lower limits are extracted from the abnormal change point and compared with the preset optimal operating range [L, U], where L represents the lower limit of the optimal operating range and U represents the upper limit. Specifically, L and U are set based on historical operating data, production target requirements, and the equipment design parameter range of the operating state parameters. For example, in industrial production, historical operating temperature data ranges from [55°C to 90°C]. Analysis reveals that the equipment's optimal performance range is [60°C to 85°C]. Therefore, L is set to 60°C and U to 85°C. This setting is obtained through statistical analysis of multiple historical operating data segments. For example, based on the equipment's long-term operating performance and production quality data, a frequency distribution is fitted using kernel density estimation methods to extract the corresponding high-probability distribution interval. The upper and lower limits of this interval are then adjusted appropriately based on production requirements to establish the optimal operating range. By comparing the fluctuation range of each abnormal change point to see if it is completely within the optimal range, if the upper limit of the fluctuation range exceeds U, it is marked as "high" and the excess value is recorded. If the lower limit of the fluctuation range is lower than L, it is marked as "low" and the below value is recorded. Suppose the abnormal fluctuation information records that the upper and lower limits of the temperature parameter fluctuation range of a node are [58°C, 88°C], and the preset optimal operating range is [60°C, 85°C]. The comparison shows that the lower limit of the temperature parameter fluctuation range is 60°C lower than the lower limit of the optimal range, which is 60-58 = 2°C lower than the lower limit of the optimal range. At the same time, the upper limit of the fluctuation range is 85°C higher than the upper limit of the optimal range, which is 88-85 = 3°C higher than the upper limit of the optimal range. The optimized range of the abnormal operating state parameter is obtained as the range between below 2°C and above 3°C. The optimization goal is to adjust the temperature parameter fluctuation range to the optimal operating range of [60°C, 85°C].

[0122] By referring to the changing trends of the time series corresponding to different operating status parameters, the correlation strength between the optimization interval of abnormal operating status parameters and the entire product production node is analyzed to generate cross-stage parameter optimization analysis results;

[0123] Analyze the strength of the correlation between the optimization interval of abnormal operating state parameters and the entire product production node. By extracting the deviation range of the optimization interval of abnormal operating state parameters and combining it with the time series data of other relevant parameters in the product production node for analysis, focus on evaluating the correlation between different operating state parameters. The specific method includes extracting the optimization interval of the operating state parameters (such as temperature, pressure, and flow) in the time period corresponding to the abnormal change point and comparing it with the variation range of the relevant parameters of the adjacent nodes, and judging the correlation based on the fluctuation trend of the time series of the two. For example, the abnormal optimization interval of the temperature parameter of a certain node is determined to be [-2℃, +3℃], which means that the temperature fluctuation range is between below 2℃ and above 3℃, which exceeds the optimal range. Combining the operating range of the node's pressure parameter (2.3 MPa, 2.7 MPa) and flow rate parameter (100 L / s, 130 L / s), it was found that the temperature parameter's variation within the abnormal range fluctuates synchronously with the upward trend of the pressure parameter (e.g., pressure increases from 2.3 MPa to 2.8 MPa), while exhibiting an anti-correlation with the downward trend of the flow parameter (e.g., flow decreases from 120 L / s to 100 L / s). Further comparison of the operating parameters of adjacent nodes revealed that the abnormal deviation of the temperature parameter at this node exhibited a similar trend with the pressure adjustment parameters of adjacent nodes within the time period synchronized with the pressure variation, indicating a correlation between temperature and pressure within this range. Based on these analysis results, a comprehensive analysis of the relationship between temperature, pressure, and flow rate reveals that temperature and pressure fluctuate synchronously at multiple time points and exhibit a stable correlation within the time period of the marked abnormal change points. Therefore, it is believed that temperature optimization has a direct impact on pressure. However, the opposite trend between temperature and flow rate indicates that flow adjustment may be influenced by other factors or is only indirectly related to temperature. Generate cross-stage parameter optimization analysis results.

[0124] See also Figure 8 , the specific steps for obtaining the node parameter optimization decision report are:

[0125] According to the correlation strength of all operating status parameters in the cross-stage parameter optimization analysis results, the importance of different operating status parameters is determined, and the priority of the operating status parameter adjustment sequence is assigned according to the importance to generate the operating status parameter priority adjustment result;

[0126] Based on the correlations between temperature, pressure, and flow identified in the analysis, and in conjunction with the specific optimization requirements for cross-stage parameters, we extracted criteria for ranking synchronization and importance within the strength of the correlations. Prioritizing target parameters that exhibited a high degree of synchronization with other parameters during multiple periods of abnormal change. A comprehensive analysis of results showing strong synchronous fluctuations between temperature and pressure over a long time span determined that temperature adjustment had a higher priority than pressure. However, time series with inverse trends in flow were less synchronized, indicating that flow had the lowest priority. Ultimately, the temperature adjustment range and order were determined based on priority.

[0127] The results of the priority adjustment of the operating status parameters are displayed in a visual way, and a node parameter optimization decision report is generated;

[0128] Integrate clear target operating status parameter priority information into visual content, including generating priority sorting tables and trend analysis charts. The priority sorting tables are arranged in order of priority of temperature, pressure, and flow, combined with the specific fluctuation range and adjustment range of each parameter in the time series. At the same time, the adjustment range in the cross-stage parameter optimization analysis results is marked in the trend analysis chart, and the fluctuation range of each parameter is distinguished by different colors or line styles, and historical data and optimization recommendation values ​​are displayed in broken lines or bar charts. Furthermore, the parameter adjustment sequence and the corresponding adjustment period are summarized as text descriptions, and a complete optimization decision report is formed in conjunction with the trend analysis chart. The report is presented in a clear combination of text and graphics, providing a reference basis for real-time optimization and adjustment of production nodes.

[0129] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. An intelligent optimization and real-time decision-making method for industrial big data, characterized by: The method comprises: The node data time distribution analysis module obtains the operating status parameters of product production nodes from industrial big data and converts them into time series. It extracts the characteristic values ​​of the operating status parameters in each time period of the time series, counts the frequency of occurrence of the characteristic values ​​in the corresponding time period, and generates node time period distribution information. The node distribution dynamic evolution module performs dynamic trend evolution on the time series corresponding to the operating status parameters based on the node time period distribution information, generates dynamic trend evolution results, marks the time points in the dynamic trend evolution results when the operating status parameters change by more than a preset change threshold, and generates node distribution change point information; The node fluctuation range assessment module analyzes the time series corresponding to the operating status parameters adjacent to the marked time points in the node distribution change point information, determines the upper and lower limits of the fluctuation range of the target operating status parameters based on the analysis results, marks abnormal change points based on the upper and lower limits of the fluctuation range, and generates node abnormal fluctuation information; The node multi-stage association optimization module obtains the abnormal operation state parameter optimization interval from the upper and lower limits of the fluctuation range of the target operation state parameter corresponding to the abnormal change point in the abnormal fluctuation information of the node and the optimal operation range of the preset operation state parameter, analyzes the correlation strength between the abnormal operation state parameter optimization interval and the entire product production node, and generates a cross-stage parameter optimization analysis result; The node parameter suggestion generation module determines the adjustment order of different operating status parameters according to the correlation strength of all operating status parameters in the cross-stage parameter optimization analysis results, and generates a node parameter optimization decision report according to the adjustment order; The steps for obtaining the node time period distribution information are specifically as follows: Obtain the operating status parameters of product production nodes from industrial big data and convert them into time series. Divide the time series into different time periods and obtain the characteristic values ​​of the operating status parameters in each time period. The characteristic values ​​include maximum value, minimum value and quantile value, and generate the characteristic value set of the operating status parameters. Counting the frequency of occurrence of each characteristic value in the set of characteristic values ​​of the operating status parameters within the corresponding time period, constructing a histogram based on the frequency of occurrence to display the frequency distribution of the operating status parameters, and using the kernel density estimation method to perform smooth fitting on the frequency distribution to generate node time period distribution information; The steps for obtaining the dynamic trend evolution results are specifically as follows: Based on the node time period distribution information, the formula is adopted: In the time series corresponding to the calculation of the operating state parameters, the operating state parameters at the previous moment are S t-1 When the current operating state parameter is S t The conditional probability P(S t |S t-1 ), generate the state change probability of the operating state parameters; Among them, A(S t-1 ,S t ) is obtained from the statistics of the number of changes of the operating state parameters in the time series corresponding to the operating state parameters, and is the operating state parameter S at the previous moment. t-1 Transfer to the current operating state parameter S t The probability of n is the total number of all operating state parameters, B(S t ,O t ) is obtained based on the frequency of occurrence of each characteristic value in the operating state parameter characteristic value set in the corresponding time period in the current operating state parameter S t The specified operating state parameter O is observed t The observation probability of k represents the index of all operating state parameters; According to the state change probability of the operating state parameters, the change rules of the operating state parameters in different periods are analyzed to generate dynamic trend evolution results.

2. The intelligent optimization and real-time decision-making method for industrial big data according to claim 1 is characterized in that: The steps for obtaining the node distribution change point information are specifically as follows: Based on the dynamic trend evolution result, by calculating the change amount of the operating status parameter in the time series corresponding to the operating status parameter point by point, marking the time point when the change amount exceeds the preset change threshold, wherein the marked time point represents the moment when a key change occurs in the time series corresponding to the operating status parameter, and generating a marked time point set; The change types and distribution density corresponding to the operating status parameters in the marked time point set are classified and counted, and the distribution law of each marked time point is analyzed in combination with the time interval of the time point to generate node distribution change point information.

3. The intelligent optimization and real-time decision-making method for industrial big data according to claim 1 is characterized in that: The steps of determining the upper and lower limits of the fluctuation range of the target operating state parameter are specifically as follows: The time series corresponding to the running state parameters adjacent to the marked time points in the node distribution change point information are analyzed using the formula: Calculate the speed v at which the target operating state parameter deviates from the current value x to the corresponding mean μ and the fluctuation amplitude σ of the target operating state parameter respectively; Among them, k' represents the intensity of the target operating state parameter returning to the mean after deviating from the corresponding mean, is the proportional coefficient, x is the current value of the target operating state parameter, μ represents the mean corresponding to the target operating state parameter, is the instantaneous change rate of the adjacent operating state parameters, β is the instantaneous change rate weight coefficient set according to the instantaneous change rate, D represents the basic value of the fluctuation range of the target operating state parameter, Δx is the fluctuation increment, which represents the difference between the maximum and minimum values ​​in the time series corresponding to the operating state parameter, and γ is the fluctuation range adjustment coefficient set according to the fluctuation increment; According to the speed v at which the target operating state parameter deviates from the current value x to the corresponding mean μ and the fluctuation amplitude σ of the target operating state parameter, the formula is adopted: [x] min ,x max ]=[μ-(σ+v·Δt),μ+(σ+v·Δt)] Determine the upper limit x of the fluctuation range of the target operating state parameter max and the lower limit x min ; Wherein, Δt represents the difference between adjacent time points in the time series corresponding to the operating state parameters.

4. The intelligent optimization and real-time decision-making method for industrial big data according to claim 3 is characterized in that: The specific steps for obtaining the abnormal node fluctuation information are as follows: Comparing the upper and lower limits of the fluctuation range of the target operating state parameter with the operating state parameter adjacent to the marked time point, marking abnormal change points exceeding the upper and lower limits of the fluctuation range according to the comparison results, and generating abnormal change point marking results; Based on the abnormal change point marking results, the distribution characteristics of all abnormal change points are classified and sorted to generate node abnormal fluctuation information.

5. The intelligent optimization and real-time decision-making method for industrial big data according to claim 1 is characterized in that: The steps for obtaining the node parameter optimization decision report are as follows: Determining the importance of different operating status parameters based on the correlation strength of all operating status parameters in the cross-stage parameter optimization analysis results, allocating operating status parameter adjustment order priorities based on the importance, and generating operating status parameter priority adjustment results; The operation status parameter priority adjustment result is displayed in a visual manner to generate a node parameter optimization decision report.

Images