A method for analyzing the trend of industrial time series data

Abstract

The present invention discloses a method for judging the trend of industrial time series data, including collecting industrial data within this time period, and verifying and preprocessing the industrial data to obtain an industrial time series data set; extracting features from the industrial time series data set to obtain feature data of the industrial time series data set; combining the number of interval segments preset by the target user to form multiple data intervals; approximating the industrial time series data, the approximate value is the upper limit or lower limit of the data interval in which it is located, to form an approximate value data set; combining the size relationship of the adjacent data before and after the approximate value data set to obtain a linear state coding data set, and judging the data trend according to the adjacent data in the adjacent linear coding data set. When using this method, the target user only needs to preset the number of interval segments, and the accuracy of trend judgment can be controlled by the number of interval segments, reducing the influence of data noise on the trend judgment result, and increasing the stability of the trend judgment of this method.

Description

Technical Field

[0001] The invention belongs to the field of electrical digital data processing, and in particular relates to a method for analyzing and judging the trend of industrial time series data. Background Art

[0002] In the analysis of industrial time series data, the application of trend analysis is particularly critical, because the upward or downward trend in industrial time series data usually represents a change in working conditions or an abnormal state. Industrial time series data analysis can help us understand and predict the changing trends and laws in the industrial production process.

[0003] The trend analysis of industrial time series data can be divided into four main types: upward trend, downward trend, horizontal trend and oscillating trend. However, the existing trend analysis methods of industrial time series data can usually only identify the upward trend, downward trend or horizontal trend when analyzing the trend of industrial time series data, and lack effective means to identify the oscillating trend.

[0004] Chinese patent CN202111192904.5 proposes a data trend detection method, which divides historical data into chronological order and performs curve fitting to obtain the slope, and then uses distribution symmetry detection to accurately and quickly determine the overall trend of the data. However, trend judgment depends on multiple preset parameters (such as thresholds). This method requires high professional knowledge and experience of the target user, which not only increases the difficulty for the target user to use this method for trend analysis, but also reduces the stability of this method for trend analysis. Different target users may have different trend prediction results for the same set of industrial time series data. Summary of the invention

[0005] The present invention provides a method for trend analysis of industrial time series data to solve the problems of lack of effective identification means for oscillation trends and misjudgment of trend analysis results, as well as increased operating difficulty and poor stability of trend analysis results caused by setting parameters.

[0006] The technical solution adopted by the present invention is:

[0007] A method for analyzing and judging the trend of industrial time series data, comprising:

[0008] S1: according to the set start time and end time, collect industrial data within this time period, and verify and preprocess the collected industrial data to obtain an industrial time series data set;

[0009] S2: extracting features from the industrial time series data set to obtain feature data of the industrial time series data set, wherein the feature data includes a maximum value and a minimum value in the industrial time series data set; and forming multiple data intervals in combination with the number of interval segments preset by the target user;

[0010] S3: Processing the industrial time series data set, and sequentially approximating the industrial time series data according to the data interval in which the data in the industrial time series data set is located, wherein the approximate value is the upper limit value or the lower limit value of the data interval, so as to form an approximate value data set;

[0011] S4: According to preset rules, combined with the size relationship between the adjacent data before and after the approximate value data set, a linear state coding data set is obtained, and the data trend is determined according to the adjacent data in the adjacent linear coding data set. The data trend includes an upward trend, a downward trend, a stable trend, and an oscillating trend.

[0012] In step S4, according to the preset rules, combined with the size relationship between the adjacent data in the approximate value data set, a linear state coding data set is obtained, specifically:

[0013] Establishing an initial array of the linear state coding data set and setting an initial value, wherein the data volume of the initial array is one less than the data volume of the approximate value data set;

[0014] According to the numerical size relationship of adjacent data in the approximate value data set, the initial array is assigned values ​​to form the linear state coding data set.

[0015] According to the numerical size relationship of adjacent data in the approximate value data set, the initial array is assigned values, specifically:

[0016] If the values ​​of adjacent data in the approximate value data set are equal, the corresponding data in the initial array are assigned a value of 0;

[0017] If the value of the next adjacent data in the approximate value data set is larger, the corresponding data in the initial array is assigned a value of 1;

[0018] If the previous data value of the adjacent data in the approximate value data set is larger, the corresponding data in the initial array is assigned a value of -1.

[0019] The data trend is determined based on the adjacent data in the adjacent linear coding data set, specifically:

[0020] If the product of adjacent data in the adjacent linear coding data set is -1, it is determined that the corresponding two intervals in the industrial time series data set are oscillating trends;

[0021] If the product of adjacent data in the adjacent linear coding data set is not -1, it is determined that the previous interval of the two corresponding intervals in the industrial time series data set is a linear trend, and the linear trend includes an upward trend, a downward trend and a stable trend.

[0022] If the product of adjacent data in the adjacent linear coding data set is not -1, it is determined that the first interval of the two corresponding intervals in the industrial time series data set is a linear trend, and the upward trend, the downward trend, and the stable trend are analyzed and judged according to the preset rules, specifically:

[0023] If the previous data of the adjacent data in the adjacent linear coding data set is 1, it is determined that the previous interval of the two corresponding intervals in the industrial time series data set is an upward trend;

[0024] If the previous data of the adjacent data in the adjacent linear coding data set is -1, it is determined that the previous interval of the two corresponding intervals in the industrial time series data set is a downward trend;

[0025] If the previous data of the adjacent data in the adjacent linear coding data set is 0, it is determined that the previous interval of the two corresponding intervals in the industrial time series data set is a stable trend.

[0026] A method for analyzing and judging the trend of industrial time series data, further comprising:

[0027] S5: According to the judged data trend, the time intervals corresponding to the upward trend, the downward trend, the stable trend and the oscillating trend are counted, and data statistical analysis is performed, including the time range duration corresponding to each data trend and the proportion in the corresponding time period of the industrial data.

[0028] In step S3, the approximate value is the upper limit or lower limit of the data interval to form an approximate value data set, specifically:

[0029] The approximate value is the lower limit of the data interval;

[0030] Alternatively, the approximate value is the upper limit of the interval;

[0031] Alternatively, the degree of proximity between the data in the industrial time series data set and the upper and lower limits of the corresponding data interval is determined; if the data is closer to the upper limit, the upper limit is approximately taken; otherwise, the lower limit is approximately taken.

[0032] In step S2, the number of interval segments preset by the target user is an integer and is greater than or equal to 2.

[0033] In step S1, the industrial data is verified, specifically:

[0034] Performing integrity verification on the industrial data to verify the continuity of the industrial data;

[0035] And / or, performing volatility verification on the industrial data to verify the validity of the industrial data.

[0036] In step S1, the collected industrial data is preprocessed, specifically:

[0037] If it is verified that the industrial data has data missing, interpolation processing is performed on the industrial data;

[0038] If it is verified that the industrial data has abnormal values, the industrial data is smoothed and / or the abnormal values ​​are eliminated.

[0039] Due to the adoption of the above technical solution, the beneficial effects achieved by the present invention are as follows:

[0040] 1. In the present invention, within the range of the maximum and minimum values ​​of the industrial time series data set, the data is divided into multiple data intervals by the number of interval segments preset by the target user. And according to the data interval where the data in the industrial time series data set is located, the industrial time series data is approximated, and the approximate value is the upper limit or lower limit of the data interval, so as to form an approximate value data set. According to the preset rules, combined with the size relationship of the adjacent data before and after in the approximate value data set, a linear state coding data set is obtained, and the data trend is judged according to the adjacent data in the adjacent linear coding data set.

[0041] When collecting the industrial data, due to the collection errors of the data collection sensors or the occurrence of environmental noise, the obtained industrial time series data will inevitably have data noise jitter, which will affect the accuracy of subsequent trend analysis. For example, due to data noise jitter, a steady trend may be misjudged as an oscillating trend, or multiple preset parameters (such as thresholds) need to be set. The preset parameters are subject to the professional knowledge and experience of the target users, and different trend analysis accuracies may be generated for different industrial time series data or different target users, which increases the workload of the target users and increases the complexity of trend analysis of industrial time series data.

[0042] The present invention can reduce the slight jitter of the collected industrial time series data caused by the collection error of the data collection sensor or the environmental noise by processing the approximate value of the industrial time series data set, thereby facilitating the trend analysis in subsequent steps and improving the accuracy of the trend analysis.

[0043] Furthermore, when the target user uses the method of the present invention to conduct trend analysis of industrial time series data, he only needs to preset the number of interval segments without inputting other parameters. The accuracy of trend analysis can be controlled by the number of interval segments. For example, the higher the number of interval segments, the higher the accuracy of trend analysis. The number of interval segments is less affected by different industrial time series data. By matching the number of interval segments with the accuracy of trend analysis, the same number of interval segments can be used for different industrial time series data to achieve the same accuracy of trend analysis. The user's identification process for industrial time series data is reduced. It is only necessary to set the number of interval segments according to the requirements of the accuracy of trend analysis, thereby improving the accuracy of the method used in the present invention. The stability of the trend analysis of this method is also increased. For the same industrial time series data, different target users can obtain the same trend analysis results, reducing the dependence on the target user's knowledge and experience.

[0044] 2. In the present invention, according to the preset rules, combined with the size relationship of the adjacent data before and after in the approximate value data set, a linear state coding data set is obtained, and the data trend is judged according to the adjacent data in the adjacent linear coding data set. In the prior art, directly judging the trend of industrial time series data requires not only to compare the size relationship of the adjacent data before and after in the industrial time series data, but also to calculate the difference between the adjacent data, and compare the difference with the preset threshold value to reduce the influence of data noise, otherwise there will be a trend misjudgment problem. That is, there are problems such as complex operation, long calculation time, and high computing resource occupation in the prior art. When judging the trend, the data in the approximate value data set is approximated, which reduces the influence of data noise, and only needs to compare the size relationship of the adjacent data before and after in the approximate value data set to obtain the linear state coding data set, and judge the data trend according to the adjacent data in the adjacent linear coding data set. The trend judgment step is simpler, the rate of trend judgment is improved, and the occupation of computing resources is reduced.

[0045] In addition, the present invention can analyze and judge multiple data trends such as upward trend, downward trend, stable trend, and oscillating trend, covering all trend patterns of industrial time series data, facilitating the identification of abnormal working conditions based on the trend of industrial time series data, and helping to promptly discover problems in the production process and make adjustments. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0047] Figure 1 A schematic diagram of a process of the industrial time series data trend analysis method according to an embodiment of the present invention;

[0048] Figure 2 A comparison diagram of an industrial time series data curve and an approximate value data curve formed after approximate value taking of industrial time series data in one embodiment of the present invention;

[0049] Figure 3 It is the time length corresponding to the four data trends after the trend analysis of industrial time series data in one embodiment of the present invention;

[0050] Figure 4 It is the time proportion corresponding to four data trends after trend analysis of industrial time series data in one implementation mode of the present invention. DETAILED DESCRIPTION

[0051] In order to more clearly illustrate the overall concept of the present invention, a detailed description is given below in an exemplary manner in conjunction with the accompanying drawings.

[0052] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0053] like Figure 1 As shown, a method for analyzing the trend of industrial time series data includes:

[0054] S1: according to the set start time and end time, collect industrial data within this time period, and verify and preprocess the collected industrial data to obtain an industrial time series data set;

[0055] This step is to process industrial time series data and analyze the trends of industrial time series data in different time intervals. The industrial data recorded by industrial sensors are generally stored in the form of industrial data and corresponding time. The time interval between two adjacent points of industrial data is the sampling interval. In order to facilitate the calculation of data, the order of data in industrial data is generally encoded in sequence, and the sampling interval is recorded separately.

[0056] In this step, the industrial data is first verified to determine the integrity and validity of the industrial data. Avoid trend analysis errors caused by incomplete industrial data, or statistical errors in the time intervals corresponding to each trend after trend analysis. In addition, by determining the validity of the industrial data, trend analysis is performed on valid industrial data to provide accurate trend analysis results, thereby accurately identifying industrial anomalies. When industrial data fails, trend analysis resources are not occupied, and the target user can be reminded to inspect and maintain the industrial data sensors. Avoid the failure of industrial data due to sensor anomalies, and avoid incorrect predictions of trends and industrial operating conditions based on failed industrial data.

[0057] Secondly, this step clears the time data, encodes the data in the order of the industrial data, and records the sampling time interval. This makes it easier for the method to traverse and loop the data, thereby facilitating the trend prediction of industrial time series data. This avoids the need to change the time interval in the method when processing different industrial time series data, which can improve the versatility of the method and reduce the number of changes to the method.

[0058] Specifically, the industrial time series data set is recorded as ,in is any data in the industrial time series dataset, for Rank order encoding in industrial time series datasets, is the total number of data in the industrial time series dataset.

[0059] It should be noted that the industrial data in the present invention includes but is not limited to parameters such as temperature, pressure, flow, speed, power, current, voltage, content, etc. By measuring and monitoring these parameters, the trend of the industrial production process can be more accurately grasped, thereby improving production efficiency, optimizing product quality, reducing energy consumption, etc.

[0060] S2: Extract features from the industrial time series data set to obtain feature data of the industrial time series data set, wherein the feature data includes the maximum value and the minimum value in the industrial time series data set; and form multiple data intervals in combination with the number of intervals preset by the target user. Figure 2 As shown, the number of target user preset interval segments is 6.

[0061] This step aims to form multiple data intervals based on the characteristic data in the industrial time series data set, that is, the maximum and minimum values ​​in the industrial time series data set, through the number of intervals preset by the target user. The method in the present invention automatically extracts the characteristic data without the need for user input.

[0062] The user only needs to preset the number of interval segments without inputting other parameters. The accuracy of trend analysis can be controlled by the number of interval segments. For example, the higher the number of interval segments, the higher the accuracy of trend analysis. The number of interval segments is less affected by different industrial time series data. By matching the number of interval segments with the accuracy of trend analysis, the same number of interval segments can be used for different industrial time series data to achieve the same accuracy of trend analysis. The user's identification process for industrial time series data is reduced. It is only necessary to set the number of interval segments according to the requirements of trend analysis accuracy, thereby improving the accuracy of the method used in the present invention. It also increases the stability of trend analysis of the method. For the same industrial time series data, different target users can obtain the same trend analysis results, reducing the dependence on the target user's knowledge and experience.

[0063] S3: Processing the industrial time series data set, and performing approximate value taking on the industrial time series data according to the data interval in which the data in the industrial time series data set is located, wherein the approximate value is the upper limit value or the lower limit value of the data interval, so as to form an approximate value data set, such as Figure 2 Data shown are approximate values.

[0064] This step aims to approximate the industrial time series data according to the multiple data intervals formed in step S2 and the data intervals in which the data in the industrial time series data set is located. The approximate value processing of the industrial time series data set can reduce the slight jitter of the collected industrial time series data due to the acquisition error of the data acquisition sensor or the environmental noise, facilitate the trend analysis in the subsequent steps, and improve the accuracy of the trend analysis.

[0065] It should be noted that, according to the data interval in which the industrial time series data is located, the value can be taken as the upper limit value of the data interval at the same time, or the value can be taken as the lower limit value of the data interval at the same time. The present invention is not limited to this.

[0066] S4: According to preset rules, combined with the size relationship between the adjacent data before and after the approximate value data set, a linear state coding data set is obtained, and the data trend is determined according to the adjacent data in the adjacent linear coding data set. The data trend includes an upward trend, a downward trend, a stable trend, and an oscillating trend.

[0067] This step aims to compare the size relationship between the adjacent data before and after the approximate value data set, obtain the linear state coding data set, and judge the data trend according to the adjacent data in the adjacent linear coding data set.

[0068] When conducting trend analysis, the data in the approximate value data set is approximated, which reduces the impact of data noise, makes the steps simpler, improves the rate of trend analysis, and reduces the occupation of computing resources. In addition, the present invention can analyze the trends of multiple data such as rising trends, falling trends, stable trends, and oscillating trends, covering all trend patterns of industrial time series data, making it easy to identify abnormal working conditions according to the trends of industrial time series data, and helping to timely discover problems in the production process and make adjustments.

[0069] As a preferred implementation of the present invention, in step S4, according to a preset rule, combined with the size relationship between the adjacent data before and after the approximate value data set, a linear state coding data set is obtained, specifically:

[0070] Establishing an initial array of the linear state coding data set and setting an initial value, wherein the data volume of the initial array is one less than the data volume of the approximate value data set;

[0071] According to the numerical size relationship of adjacent data in the approximate value data set, the initial array is assigned values ​​to form the linear state coding data set.

[0072] This step aims to establish an initial array of a linear state coding data set in order to store the linear state coding data. Each linear state coding data represents the numerical size relationship of adjacent data in the approximate value data set, so the data volume of the initial array is one less than the data volume of the approximate value data set. At the same time, it is convenient to perform trend analysis based on the relationship between adjacent data in the linear state coding data set.

[0073] Specifically, the approximate value data set is recorded as ,in, Industrial time series data Approximate value data after approximate value.

[0074] The linear state coding dataset is recorded as ,in Based on and As a preferred embodiment of this implementation, the initial array is assigned values ​​according to the numerical size relationship of adjacent data in the approximate value data set, specifically:

[0075] If the values ​​of adjacent data in the approximate value data set are equal, the corresponding data in the initial array are assigned a value of 0;

[0076] If the value of the next adjacent data in the approximate value data set is larger, the corresponding data in the initial array is assigned a value of 1;

[0077] If the previous data value of the adjacent data in the approximate value data set is larger, the corresponding data in the initial array is assigned a value of -1.

[0078] The numerical sizes of adjacent data in the approximate value data set include three situations, namely, the latter data is equal to the previous data, the latter data value is relatively large, and the previous data value is relatively large. According to these three situations, the corresponding data in the initial array are assigned values ​​of 0, 1, and -1 one by one, so as to facilitate subsequent trend analysis based on adjacent data calculations in adjacent linear coding data sets.

[0079] Specifically, if ,but Assign 0; if ,but Assign -1; if but Assign a value of 1.

[0080] It should be noted that the present invention does not limit the assignment of corresponding data in the initial array according to the three situations that the latter data is equal to the previous data, the latter data value is relatively large, and the previous data value is relatively large. The corresponding data in the initial array can be assigned to -1, 0 and 1, or 1, 2, 3 one by one. It is only necessary to adjust the corresponding rules when judging the data trend according to the adjacent data in the adjacent linear coding data set.

[0081] Preferably, the data trend is determined based on adjacent data in adjacent linear coding data sets, specifically:

[0082] If the product of adjacent data in the adjacent linear coding data set is -1, it is determined that the corresponding interval in the industrial time series data set is an oscillation trend;

[0083] If the product of adjacent data in the adjacent linear coding data set is not -1, it is determined that the previous data in the two corresponding intervals in the industrial time series data set is a linear trend, and the linear trend includes an upward trend, a downward trend and a stable trend.

[0084] It should be noted that in the approximate value data set corresponding to this step, the latter data is equal to the previous data, the latter data value is relatively large, and the previous data value is relatively large. According to these three situations, the corresponding data in the initial array are assigned 0, 1 and -1 one by one.

[0085] In this step, the oscillation trend and the linear trend are distinguished according to the product of the adjacent data in the adjacent linear coding data set. When the product of the adjacent data in the adjacent linear coding data set is -1, it means that the trend of the adjacent data of the previous industrial time series is opposite to that of the adjacent data of the next industrial time series. Specifically, the adjacent data of the previous industrial time series is an upward trend, and the adjacent data of the next industrial time series is a downward trend, or the adjacent data of the previous industrial time series is a downward trend, and the adjacent data of the next industrial time series is an upward trend. Therefore, when the product of the adjacent data in the adjacent linear coding data set is -1, it is determined to be an oscillation trend.

[0086] Specifically, hour, , , indicating that the industrial time series data set ~ The range is an upward trend. ~ The interval is a downward trend, comprehensive judgment ~ The range is an oscillating trend. Or , , indicating that the industrial time series data set ~ The range is a downward trend. ~ The interval is an upward trend, comprehensive judgment ~ The range is an oscillating trend.

[0087] It should be noted that adjacent linear coding data sets are traversed to determine the trend of industrial time series data sets. ~ When the interval is an oscillating trend, the next step is to perform t+2 operation on t, that is, through Conduct trend analysis to avoid repeated analysis of the same range.

[0088] When the initial array is assigned values ​​according to other rules, the calculation principle needs to be adjusted accordingly, which is not limited in the present invention.

[0089] Furthermore, if the product of adjacent data in the adjacent linear coding data set is not -1, it is determined that the first interval of the two corresponding intervals in the industrial time series data set is a linear trend, and the upward trend, the downward trend, and the stable trend are analyzed and judged according to preset rules, specifically:

[0090] If the previous data of the adjacent data in the adjacent linear coding data set is 1, it is determined that the previous interval of the two corresponding intervals in the industrial time series data set is an upward trend;

[0091] If the previous data of the adjacent data in the adjacent linear coding data set is -1, it is determined that the previous interval of the two corresponding intervals in the industrial time series data set is a downward trend;

[0092] If the previous data of the adjacent data in the adjacent linear coding data set is 0, it is determined that the previous interval of the two corresponding intervals in the industrial time series data set is a stable trend.

[0093] Specifically, hour, When, it means that the industrial time series data is concentrated ~ The range is an upward trend; When, it means that the industrial time series data is concentrated ~ The interval is a stable trend; When, it means that the industrial time series data is concentrated ~ The range is in a downward trend.

[0094] It should be noted that When , only the first interval of the two corresponding intervals in the industrial time series data set ( ~ The next step is to perform the t+1 operation on t, that is, Conduct trend analysis.

[0095] As a preferred embodiment of the present invention, Figure 1 As shown in the figure, the trend analysis method of industrial time series data also includes:

[0096] S5: According to the judged data trend, the time intervals corresponding to the upward trend, the downward trend, the stable trend and the oscillating trend are counted, and data statistical analysis is performed, including the time range duration corresponding to each data trend and the proportion in the corresponding time period of the industrial data.

[0097] This step aims to count the time ranges corresponding to the upward trend, downward trend, stable trend, and oscillating trend, and calculate the duration of the time range corresponding to the four data trends, as well as the proportion in the corresponding time period of industrial data. This will make it easier for target users to conduct a comprehensive analysis of trend analysis data, help identify problems in the production process, and make adjustments.

[0098] As a preferred embodiment of the present invention, the method for analyzing the trend of industrial time series data further includes: in step S3, the approximate value is the upper limit or lower limit of the data interval to form an approximate value data set, specifically:

[0099] The approximate value is the lower limit of the data interval;

[0100] Alternatively, the approximate value is the upper limit of the interval;

[0101] Alternatively, the degree of proximity between the data in the industrial time series data set and the upper and lower limits of the corresponding data interval is determined; if the data is closer to the upper limit, the upper limit is approximately taken; otherwise, the lower limit is approximately taken.

[0102] This step aims to approximate the values ​​of the industrial time series data and uniformly select the upper limit value or the lower limit value of the interval. The present invention does not impose any restrictions on this and a unified approximation rule can be adopted.

[0103] Alternatively, the degree of proximity between the data in the industrial time series data set and the upper and lower limits of the corresponding data interval is determined. If the data is closer to the upper limit, the upper limit is approximated, otherwise the lower limit is approximated. The value is taken based on the degree of proximity between the industrial time series data and the upper and lower limits of the corresponding data interval to avoid errors in trend analysis caused by large increases or decreases when approximating the industrial time series data.

[0104] As a preferred implementation of the present invention, in step S2, the number of target user preset interval segments is an integer and is greater than or equal to 2.

[0105] When the target user presets the number of interval segments to 1, the accuracy of trend analysis is reduced. That is, the industrial time series data is approximated to the maximum or minimum value, the approximation error increases, and the trend analysis results are affected. The preset number of interval segments is an integer, which makes it easier to divide the industrial time series data set into integer data intervals according to the value size, and approximate all industrial time series data with similar approximate accuracy. Avoid the increase in the approximate value error of industrial time series data in a certain data interval, which affects the trend analysis results.

[0106] The target user can control the accuracy of trend analysis through the number of interval segments. For example, the larger the number of interval segments, the higher the accuracy of trend analysis. The number of interval segments is less affected by different industrial time series data. By matching the number of interval segments with the accuracy of trend analysis, the same number of interval segments can be used for different industrial time series data to achieve the same accuracy of trend analysis. The user's identification process of industrial time series data is reduced. It is only necessary to set the number of interval segments according to the requirements of trend analysis accuracy, thereby improving the accuracy of the method used in the present invention. It also increases the stability of trend analysis of the method. For the same industrial time series data, different target users can obtain the same trend analysis results, reducing the dependence on the target user's knowledge and experience.

[0107] As a preferred implementation of the present invention, in step S1, the industrial data is verified, specifically:

[0108] Performing integrity verification on the industrial data to verify the continuity of the industrial data;

[0109] And / or, performing volatility verification on the industrial data to verify the validity of the industrial data.

[0110] Through integrity verification, verify that industrial data is continuous, without missing values ​​or gaps. Perform volatility detection on industrial data. To ensure that industrial data is obvious and effective, industrial data should show a certain volatility or variability, that is, avoid situations where data points are static or constant on the time axis (which means that industrial data is invalid and industrial data collection sensors are faulty). Volatility can be evaluated by calculating statistics of industrial data (such as variance, standard deviation). Of course, volatility can also be calculated in other ways, and there is no restriction on this.

[0111] Preferably, in step S1, the collected industrial data is preprocessed, specifically:

[0112] If it is verified that the industrial data has data missing, interpolation processing is performed on the industrial data;

[0113] If it is verified that the industrial data has abnormal values, the industrial data is smoothed and / or the abnormal values ​​are eliminated.

[0114] This step is to process abnormal industrial data to facilitate subsequent data trend analysis and avoid the impact of missing data or the existence of abnormal values ​​on the trend analysis results.

[0115] As a specific embodiment of the present invention, the transmission shaft amplitude data of a coal mine ventilator is used as a test case. The industrial data is sampled at an interval of about 5 seconds. The entire data set covers 1947 sampling points. The curve and trend analysis results after approximate value are as follows: Figures 2 to 4 The method proposed in the present invention can accurately identify the duration of each trend in industrial time series data and its relative proportion in the entire time period, so as to more objectively understand and predict the dynamic changes in the industrial system, and facilitate timely discovery of problems in the production process and make adjustments.

[0116] Anything not described in the present invention can be achieved by adopting or drawing on existing technologies.

[0117] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0118] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A method for analyzing the trend of industrial time series data, characterized in that: According to the set start time and end time, industrial data within this time period is collected, and the collected industrial data is verified and preprocessed to obtain an industrial time series data set; Perform feature extraction on the industrial time series data set to obtain feature data of the industrial time series data set, wherein the feature data includes a maximum value and a minimum value in the industrial time series data set; and form multiple data intervals in combination with the number of interval segments preset by the target user; Processing the industrial time series data set, and sequentially approximating the industrial time series data according to the data interval in which the data in the industrial time series data set is located, wherein the approximate value is an upper limit value or a lower limit value of the data interval, so as to form an approximate value data set; According to a preset rule, combined with the size relationship between the adjacent data in the approximate value data set, data values ​​are assigned one by one according to the three situations that the latter data is equal to the former data, the latter data value is relatively large, and the former data value is relatively large, to obtain a linear state coding data set; Determine data trend according to adjacent data in adjacent linear coding data sets, wherein the data trend includes oscillation trend and linear trend. When the trend of the current industrial time series adjacent data is opposite to that of the subsequent industrial time series adjacent data, it is determined to be an oscillating trend; when the trend of the current industrial time series adjacent data is the same as that of the subsequent industrial time series adjacent data, it is determined to be a linear trend, and the linear trend includes an upward trend, a downward trend, and a stable trend.

2. The method for analyzing the trend of industrial time series data according to claim 1 is characterized in that: According to the preset rules, combined with the size relationship between the adjacent data in the approximate value data set, a linear state coding data set is obtained, specifically: Establishing an initial array of the linear state coding data set and setting an initial value, wherein the data volume of the initial array is one less than the data volume of the approximate value data set; According to the numerical size relationship of adjacent data in the approximate value data set, the initial array is assigned values ​​to form the linear state coding data set.

3. The method for analyzing the trend of industrial time series data according to claim 2 is characterized in that: According to the numerical size relationship of adjacent data in the approximate value data set, the initial array is assigned values, specifically: If the values ​​of adjacent data in the approximate value data set are equal, the corresponding data in the initial array are assigned a value of 0; If the value of the next adjacent data in the approximate value data set is larger, the corresponding data in the initial array is assigned a value of 1; If the previous data value of the adjacent data in the approximate value data set is larger, the corresponding data in the initial array is assigned a value of -1.

4. The method for analyzing the trend of industrial time series data according to claim 3 is characterized in that: The data trend is determined based on the adjacent data in the adjacent linear coding data set, specifically: If the product of adjacent data in the adjacent linear coding data set is -1, it is determined that the corresponding two intervals in the industrial time series data set are oscillating trends; If the product of adjacent data in the adjacent linear coding data set is not -1, it is determined that the previous interval of the two corresponding intervals in the industrial time series data set is a linear trend, and the linear trend includes an upward trend, a downward trend and a stable trend.

5. The method for analyzing the trend of industrial time series data according to claim 4 is characterized in that: If the product of adjacent data in the adjacent linear coding data set is not -1, it is determined that the first interval of the two corresponding intervals in the industrial time series data set is a linear trend, and the upward trend, the downward trend, and the stable trend are analyzed and judged according to the preset rules, specifically: If the previous data of the adjacent data in the adjacent linear coding data set is 1, it is determined that the previous interval of the two corresponding intervals in the industrial time series data set is an upward trend; If the previous data of the adjacent data in the adjacent linear coding data set is -1, it is determined that the previous interval of the two corresponding intervals in the industrial time series data set is a downward trend; If the previous data of the adjacent data in the adjacent linear coding data set is 0, it is determined that the previous interval of the two corresponding intervals in the industrial time series data set is a stable trend.

6. The method for analyzing the trend of industrial time series data according to claim 1 is characterized in that: Also includes: According to the judged data trend, the time intervals corresponding to the upward trend, the downward trend, the stable trend and the oscillating trend are statistically analyzed, and data statistical analysis is performed, including the duration of the time range corresponding to each data trend and the proportion in the corresponding time period of the industrial data.

7. The method for analyzing the trend of industrial time series data according to claim 1 is characterized in that: The approximate value is the upper or lower limit of the data interval to form an approximate value data set, specifically: The approximate value is the lower limit of the data interval; Alternatively, the approximate value is the upper limit of the interval; Alternatively, the degree of proximity between the data in the industrial time series data set and the upper limit and lower limit of the corresponding data interval is determined. If the data is closer to the upper limit, the upper limit is approximately taken as the value; otherwise, the lower limit is approximately taken as the value.

8. The method for analyzing the trend of industrial time series data according to claim 1 is characterized in that: The number of target user preset interval segments is an integer and is greater than or equal to 2.

9. The method for analyzing the trend of industrial time series data according to claim 1, characterized in that: Verify the industrial data, specifically: Performing integrity verification on the industrial data to verify the continuity of the industrial data; And / or, performing volatility verification on the industrial data to verify the validity of the industrial data.

10. The method for analyzing the trend of industrial time series data according to claim 1, characterized in that: The collected industrial data is preprocessed, specifically: If it is verified that the industrial data has data missing, interpolation processing is performed on the industrial data; If it is verified that the industrial data has abnormal values, the industrial data is smoothed and / or the abnormal values ​​are eliminated.

Images