A clutch intelligent detection method based on multi-dimensional data acquisition

Abstract

The present invention relates to the field of data processing, and in particular to an intelligent clutch detection method based on multidimensional data acquisition, comprising: obtaining clutch-related data uploaded by a user; determining target spectral data based on a vehicle characterization coefficient; determining valid unit data based on an element characterization value, and determining a data screening method for the target spectral data based on the number of valid unit data; determining the unit data status based on a data segment reference value and a characteristic threshold; determining, based on the unit data status of the valid unit data, whether the data processing method for the valid unit data is segmentation based on concentration values ​​or segmentation based on torque thresholds; and determining, under preset conditions, whether the segmented segments are qualified based on a comprehensive deviation value. The present invention can perform targeted training on oil spectral data at different wear stages, thereby improving the accuracy of clutch aging life estimation detection.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to an intelligent clutch detection method based on multi-dimensional data acquisition. Background Art

[0002] Oil spectral data contains the concentration of wear particles of multiple elements, but not all oil spectral data can provide useful degradation characterization information. In addition, the clutch operation process includes different wear states, and the oil spectral data of different wear states are not segmented, resulting in poor accuracy in clutch life prediction. Therefore, how to conduct targeted training on oil spectral data of different wear stages to improve the accuracy of clutch aging life prediction detection is a technical problem that needs to be urgently solved by technical personnel in this field.

[0003] Chinese Patent Publication No. CN102254063B discloses a method for automatically updating a clutch database, comprising: first, determining the parts and parameter information that affect clutch performance to form an influencing parameter set; the elements in the influencing parameter set have different degrees of influence on the clutch design quality. To reflect the importance of each element, a corresponding weight should be determined for each element, forming a weight set corresponding to the influencing parameter set; then, the membership degree of the elements in the influencing parameter set is calculated using a membership function of a Cauchy distribution to construct a fuzzy relationship matrix; finally, a composite operation is performed on the influencing parameter set, the fuzzy relationship matrix, and the weight set to obtain a comprehensive evaluation value. If the value meets the requirements, the design data can be entered into the database; otherwise, the parameters must be redesigned. Therefore, the above technical solution has the following problems: it is only based on the complete oil spectral data for entry, without segmenting the oil spectral data, and cannot be targeted for training oil spectral data at different wear stages, resulting in poor accuracy in clutch aging life prediction detection. Summary of the Invention

[0004] To this end, the present invention provides an intelligent clutch detection method based on multi-dimensional data acquisition to overcome the problems in the prior art that the data is only input based on the complete oil spectral data, the oil spectral data is not segmented, and the oil spectral data of different wear stages cannot be trained in a targeted manner, resulting in poor accuracy in clutch aging life estimation detection.

[0005] To achieve the above objectives, the present invention provides a clutch intelligent detection method based on multi-dimensional data acquisition, comprising:

[0006] Get clutch related data uploaded by users;

[0007] determining target spectral data according to the vehicle characterization coefficient;

[0008] Determine the valid unit data according to the element characterization value, and determine the target spectrum data according to the number of valid unit data. The data screening method is to screen according to the correlation coefficient, or to screen Ф valid unit data in descending order according to the element characterization value;

[0009] Determine the unit data state according to the data segment reference value and the feature threshold;

[0010] Determining, according to the unit data status of the valid unit data, a data processing method of the valid unit data, that is, segmenting according to a concentration value, or segmenting according to a torque threshold;

[0011] Under the preset conditions, whether the segmented paragraph is qualified is determined based on the comprehensive deviation value;

[0012] The preset condition is that the segmentation is completed according to the torque threshold.

[0013] Further, target spectral data is determined according to the vehicle characterization coefficient;

[0014] Dividing the data sequence of the spectral data and obtaining a number of pre-screened spectral data, each pre-screened spectral data being recorded as target spectral data;

[0015] The data sequence of the spectral data is in descending order of vehicle characterization coefficients;

[0016] The vehicle characterization coefficient is determined according to the block frequency reference value and the environment characterization value;

[0017] The relationship between the vehicle characterization coefficient and the frequency blocking reference value is a positive correlation, and the relationship between the vehicle characterization coefficient and the environment characterization value is a positive correlation.

[0018] Further, valid unit data is determined according to the element characterization value, and a data screening method of the target spectrum data is determined according to the number of valid unit data;

[0019] If the number of valid unit data is greater than or equal to the preset number, the data screening method is to filter according to the correlation coefficient;

[0020] If the number of valid unit data is less than the preset number, the data screening method is to screen Ф valid unit data in descending order according to the element representation value;

[0021] The valid unit data is unit data whose element representation value is greater than a preset element representation value.

[0022] Furthermore, screening based on correlation coefficients includes:

[0023] For a target spectral data, the valid unit data with the largest element characterization value in the target spectral data is recorded as the target unit, and the other valid unit data in the target spectral data excluding the target unit is recorded as the reference unit. The correlation coefficient between the target unit and each reference unit is detected, and Φ-1 reference units are screened in descending order of the correlation coefficient. The target unit and the screened Φ-1 reference units are recorded as screened data.

[0024] Further, determining a data processing mode of the valid unit data according to the unit data status of the valid unit data;

[0025] If the unit data state is the first preset unit data state, segmentation is performed according to the concentration value;

[0026] If the unit data state is the second preset unit data state, segmentation is performed according to the torque threshold.

[0027] Furthermore, the unit data state is determined based on the data segment reference value and the characteristic threshold, and the unit data state includes:

[0028] A first preset unit data state in which the data segment reference value is less than a preset data segment reference value and the characteristic threshold is less than a preset characteristic threshold;

[0029] A second preset unit data state in which the data segment reference value is greater than or equal to the preset data segment reference value or the feature threshold is greater than or equal to the preset feature threshold;

[0030] The data segment reference value is determined according to the number of characteristic segments and the maximum slope deviation, and the characteristic threshold is determined according to the data increase and the unit data duration.

[0031] Furthermore, segmentation is performed based on concentration values;

[0032] For a valid unit data, the starting concentration point and the ending concentration point of the data segment whose concentration value is greater than the first preset concentration and less than the second preset concentration are used as segmentation points for segmentation;

[0033] The first preset concentration is less than the second preset concentration.

[0034] Furthermore, segmentation is performed based on a torque threshold;

[0035] For torque data corresponding to a valid unit of data, a torque segment having a torque reference coefficient greater than or equal to a preset torque reference coefficient is recorded as a characteristic torque segment, and distribution detection is performed on each characteristic torque segment. When distribution detection is performed on a single characteristic torque segment, the characteristic torque segment is recorded as a target characteristic torque segment. A distance reference value between each characteristic torque segment after the time sequence of the target characteristic torque segment is detected. If the distance reference value between each characteristic torque segment after the time sequence of the target characteristic torque segment is less than the preset distance reference value, the segment is recorded as an associated characteristic torque segment of the target characteristic torque segment. If the distance reference value between each characteristic torque segment after the time sequence of the target characteristic torque segment is greater than or equal to the preset distance reference value, distribution detection for the target characteristic torque segment is stopped, and the segment between the starting torque point of the target characteristic torque segment and the ending torque point of the associated characteristic torque segment with the largest distance reference value corresponding to the target characteristic torque segment is recorded as a segment combination.

[0036] The valid unit data is segmented using the starting torque point and the ending torque point corresponding to the paragraph combination with the largest horizontal coordinate length as segmentation points;

[0037] The torque threshold is determined according to a torque reference coefficient and a torque fluctuation coefficient. The torque reference coefficient is positively correlated with the torque mean value, and the torque fluctuation coefficient is positively correlated with the torque fluctuation value.

[0038] Further, determining whether the segmented paragraph is qualified based on the comprehensive deviation value;

[0039] Detecting a starting temperature point and an ending temperature point of a temperature segment where the temperature reference value is greater than or equal to a preset temperature reference value, and determining a comprehensive deviation value based on the sum of the absolute value of the difference between the starting temperature point and the starting torque point and the absolute value of the difference between the ending temperature point and the ending torque point;

[0040] If the comprehensive deviation value is greater than or equal to the preset comprehensive deviation value, the segmentation fails, and the starting midpoint and the ending midpoint are selected as the segmentation points for segmentation;

[0041] If the comprehensive deviation value is less than the preset comprehensive deviation value, the segmentation is qualified.

[0042] Furthermore, the element characterization value is determined according to the sum of the unit characterization values ​​corresponding to each data point.

[0043] Compared with the prior art, the beneficial effect of the present invention lies in that, in the technical solution of the present invention, the target spectral data is determined according to the vehicle characterization coefficient, and the vehicle characterization coefficient is used to reflect the influence of operating habits on the pre-screening spectral data, thereby dividing the data sequence of the spectral data and obtaining a number of pre-screening spectral data, avoiding the problem of poor representativeness of the target spectral data, ensuring the richness of the target spectral data, and thus improving the accuracy of the clutch aging life estimation detection.

[0044] Furthermore, the present invention effectively reflects the information content contained in the target spectral data through the element standard value, and then determines the effective unit data according to the element characterization value, so that the selection of effective unit data is more representative, so that the screened effective unit data can better characterize the wear state of the clutch, thereby improving the accuracy and reliability of the clutch aging life prediction detection.

[0045] Furthermore, the present invention determines the data screening method of the target spectral data according to the number of valid unit data, and effectively reflects the richness of the data through the number of valid unit data, and then adaptively selects different data screening methods, so that the selection of data screening method is more in line with the actual application scenario, ensuring that high-quality data can be selected when the number of valid unit data is different, avoiding the problem of inaccurate clutch life prediction due to data quality problems, and thus improving the reliability of clutch aging life estimation detection.

[0046] Furthermore, the present invention can effectively reflect the relevant situation of the effective unit data through the correlation coefficient, and then screen out the effective unit data that has an important impact on clutch wear based on the correlation coefficient, thereby improving the reliability of the clutch aging life prediction detection.

[0047] Furthermore, the present invention determines the unit data state based on the data segment reference value and the characteristic threshold, and effectively reflects the distribution points of the valid unit data through the data segment reference value and the characteristic threshold, and then selects different data processing methods according to the unit data state, so that the selection of data processing method is more in line with the actual application scenario, and can effectively divide the different wear stages, and can perform targeted training on the oil spectral data of different wear stages, thereby improving the accuracy of the clutch aging life estimation detection.

[0048] Furthermore, the present invention effectively reflects the wear state of the clutch through the concentration value, and can then divide different wear stages according to the concentration value, and can perform targeted training on the oil spectral data of different wear states, thereby improving the accuracy of clutch aging life prediction detection.

[0049] Furthermore, the present invention performs segmentation based on the torque threshold, effectively reflects the wear state of the clutch through the torque threshold, and then segments the effective unit data according to the wear state, so as to conduct targeted training on the oil spectral data of different wear states, thereby improving the accuracy of the clutch aging life prediction detection, effectively reflects the wear state of the clutch through the temperature fluctuation value, and then effectively reflects the deviation between the temperature point and the torque point through the comprehensive deviation value, and then determines whether the segmentation segment is qualified according to the actual situation, so that the setting position of the segmentation point is more accurate, thereby improving the accuracy of the clutch aging life prediction detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 Schematic diagram of the clutch intelligent detection method based on multi-dimensional data acquisition of the present invention;

[0051] Figure 2 A flow chart of a data screening method for determining target spectral data for the number of valid unit data of the present invention;

[0052] Figure 3 This is a flow chart of a data processing method for determining valid unit data according to the unit data status of the valid unit data according to the present invention;

[0053] Figure 4 This is a flow chart of the present invention for determining the unit data state according to the data segment reference value and the characteristic threshold. DETAILED DESCRIPTION

[0054] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0055] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0056] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside", and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0057] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0058] See also Figures 1 to 4 As shown, the present invention provides a clutch intelligent detection method based on multi-dimensional data acquisition, comprising:

[0059] Get clutch related data uploaded by users;

[0060] determining target spectral data according to the vehicle characterization coefficient;

[0061] Determine the valid unit data according to the element characterization value, and determine the target spectrum data according to the number of valid unit data. The data screening method is to screen according to the correlation coefficient, or to screen Ф valid unit data in descending order according to the element characterization value;

[0062] Determine the unit data state according to the data segment reference value and the feature threshold;

[0063] Determining, according to the unit data status of the valid unit data, a data processing method of the valid unit data, that is, segmenting according to a concentration value, or segmenting according to a torque threshold;

[0064] Under the preset conditions, whether the segmented paragraph is qualified is determined based on the comprehensive deviation value;

[0065] The preset condition is that the segmentation is completed according to the torque threshold.

[0066] The application scenario of the present invention is for clutch aging life estimation detection, specifically spectral data sample selection, clutch-related data includes but is not limited to spectral data, torque data and temperature data, wherein the monitoring time of the spectral data, torque data and temperature data corresponding to a single clutch-related data is the same, and a single spectral data includes a number of unit data, and the unit data is the concentration value of a single element in the clutch lubricating oil monitored at different time points during the historical use of the clutch. The lubricating oil is the oil between the clutch friction plate and the steel plate. The method for detecting the concentration value is to obtain the concentration value by measuring the oil through an oil analysis spectrometer after ultra-high voltage excitation between the disk and rod electrodes. This is content that is easy for technicians in this field to understand and will not be described in detail.

[0067] Among them, a unit data monitoring method is provided, in which lubricating oil is sampled once every 1 hour, and the time of a single extraction of lubricating oil is recorded as a time point to obtain the concentration value of the element corresponding to the unit data at each time point during the monitoring time. The torque data and temperature data are monitored by a torque sensor and a temperature sensor to obtain the torque value and temperature value corresponding to each time point during the monitoring time. This is content that is easy for technical personnel in this field to understand and will not be elaborated on in detail.

[0068] Specifically, target spectral data is determined based on the vehicle characterization coefficient;

[0069] Dividing the data sequence of the spectral data and obtaining a number of pre-screened spectral data, each pre-screened spectral data being recorded as target spectral data;

[0070] The data sequence of the spectral data is in descending order of vehicle characterization coefficients;

[0071] The vehicle characterization coefficient is determined according to the block frequency reference value and the environment characterization value;

[0072] The relationship between the vehicle characterization coefficient and the frequency blocking reference value is a positive correlation, and the relationship between the vehicle characterization coefficient and the environment characterization value is a positive correlation.

[0073] The pre-screening spectral data is obtained by dividing the spectral data sequence into M equal parts, and recording the spectral data at each equal division point as the pre-screening spectral data. The value of M is positively correlated with the length of the data sequence, and the length of the data sequence is the number of spectral data in the data sequence.

[0074] For a pre-screened spectrum data, the frequency reference value = the number of gear shifts / data duration, the number of gear shifts is the number of gear shifts within the monitoring time corresponding to the pre-screened spectrum data, the data duration is the length of the monitoring time corresponding to the pre-screened spectrum data, the environmental characterization value = (vehicle average speed / preset vehicle average speed) × speed coefficient + (vehicle setback number / preset vehicle setback number) × setback coefficient, vehicle average speed = the distance traveled by the vehicle within the monitoring time corresponding to the pre-screened spectrum data / data duration, the number of vehicle setbacks is the number of times the speed change of the vehicle within the monitoring time corresponding to the pre-screened spectrum data is greater than the preset speed change. The present invention is provided with a continuous cycle monitoring Cycle, the status is determined once at the end of each monitoring cycle. The duration of the monitoring cycle can be set according to user needs. The greater the user's demand for speed monitoring accuracy, the shorter the monitoring cycle. A value for the duration of a monitoring cycle is provided. The duration of a monitoring cycle is 1 minute. Speed ​​change = |Distance traveled by the vehicle in the current monitoring cycle / Duration of the monitoring cycle - Distance traveled by the vehicle in the previous monitoring cycle adjacent to the current monitoring cycle / Duration of the monitoring cycle|. The value of the preset speed change can be determined by the user according to the actual application scenario. A value for the preset speed change is provided. The preset speed change is 30km / h.

[0075] It is understandable that the values ​​of the speed coefficient and the jerking coefficient can be obtained by the user through deep learning convolutional neural network learning based on historical records. It is understandable that the present invention reflects the impact of vehicle driving on effective unit data through environmental characterization values. The user can use deep learning through historical user data to obtain the impact of the average vehicle speed and the number of vehicle jerking times on the effective unit data, and then select the corresponding values ​​of the speed coefficient and the jerking coefficient, where the speed coefficient + the jerking coefficient = 1. One value of the speed coefficient and the jerking coefficient is provided, and the speed coefficient is 0.6 and the jerking coefficient is 0.4.

[0076] The values ​​of the preset vehicle average speed and the preset number of vehicle jerks can be determined by the user according to the actual application scenario. The values ​​of the preset vehicle average speed and the preset number of vehicle jerks are provided. The preset vehicle average speed is the average value of the vehicle average speeds corresponding to each valid unit data, and the preset number of vehicle jerks is the average value of the vehicle jerks corresponding to each valid unit data.

[0077] Specifically, valid unit data is determined according to the element characterization value, and a data screening method of the target spectral data is determined according to the number of valid unit data;

[0078] If the number of valid unit data is greater than or equal to the preset number, the data screening method is to filter according to the correlation coefficient;

[0079] If the number of valid unit data is less than the preset number, the data screening method is to screen Ф valid unit data in descending order according to the element representation value;

[0080] The valid unit data is unit data whose element representation value is greater than a preset element representation value.

[0081] Among them, the number of valid unit data is the number of valid unit data in a target spectral data. The preset number and the preset element characterization value can be determined by the user according to the actual application scenario. The larger the values ​​of the preset number and the preset element characterization value, the more representative the target spectral data is in reflecting the clutch wear condition, and thus the better the accuracy of the clutch performance degradation model prediction. A preset number and a preset element characterization value are provided, and the preset number is 10. The method for confirming the preset element characterization value is to record the average value of the element characterization value corresponding to the valid unit data that can meet the user's needs in the historical data processing process as the preset element characterization value.

[0082] The value of Ф can be determined by the user according to the actual application scenario. The higher the user's demand for the accuracy of the clutch performance degradation model prediction, the larger the value of Ф. One value of Ф is provided, and the preset number is 4.

[0083] The calculation formula for the correlation coefficient r of two valid unit data is:

[0084]

[0085] Where n is the number of time points within the monitoring time corresponding to a single valid unit data; and are the concentration values ​​corresponding to the i-th time point during the monitoring time corresponding to the two valid unit data, for The average value of the concentration values ​​corresponding to each time point during the monitoring time corresponding to the valid unit data, for The average value of the concentration values ​​corresponding to each time point within the monitoring time corresponding to the corresponding valid unit data, i = 1, 2, 3, ..., n.

[0086] Specifically, screening based on correlation coefficients includes:

[0087] For a target spectral data, the valid unit data with the largest element characterization value in the target spectral data is recorded as the target unit, and the other valid unit data in the target spectral data excluding the target unit is recorded as the reference unit. The correlation coefficient between the target unit and each reference unit is detected, and Φ-1 reference units are screened in descending order of the correlation coefficient. The target unit and the screened Φ-1 reference units are recorded as screened data.

[0088] Specifically, the data processing mode of the valid unit data is determined according to the unit data status of the valid unit data;

[0089] If the unit data state is the first preset unit data state, segmentation is performed according to the concentration value;

[0090] If the unit data state is the second preset unit data state, segmentation is performed according to the torque threshold.

[0091] Specifically, the unit data state is determined based on the data segment reference value and the feature threshold, and the unit data state includes:

[0092] A first preset unit data state in which the data segment reference value is less than a preset data segment reference value and the characteristic threshold is less than a preset characteristic threshold;

[0093] A second preset unit data state in which the data segment reference value is greater than or equal to the preset data segment reference value or the feature threshold is greater than or equal to the preset feature threshold;

[0094] The data segment reference value is determined according to the number of characteristic segments and the maximum slope deviation, and the characteristic threshold is determined according to the data increase and the unit data duration.

[0095] The data segment reference value = (number of characteristic segments / preset number of characteristic segments) × segment coefficient + (maximum slope deviation / preset maximum slope deviation) × slope coefficient. The number of characteristic segments is determined by plotting a spectral image corresponding to a valid unit of data, with monitoring time as the abscissa and concentration value as the ordinate. The spectral image is divided into several spectral segments with the same abscissa length. Association segment analysis is performed on each spectral segment according to the chronological order of the spectral image. When performing association segment analysis on a single spectral segment, the spectral segment is recorded as a target spectral segment. The slope difference between each spectral segment and the target spectral segment after the chronological order of the target spectral segment is detected. If the slope difference between the spectral segment and the target spectral segment is less than the preset slope difference, the spectral segment is recorded as an associated spectral segment of the target spectral segment. If the slope difference between the spectral segment and the target spectral segment is greater than or equal to the preset slope difference, the association segment analysis for the target spectral segment is terminated, and the target spectral segment and its corresponding associated spectral segments are collectively recorded as one associated segment. The number of associated segments and the number of spectral segments not recorded as associated segments is recorded as the number of characteristic segments. For two spectral segments, the slope difference is equal to the absolute value of the difference between the slopes of one segment and the other. For a single spectral segment, the concentration corresponding to the leftmost time point of the segment is recorded as the first concentration, and the concentration corresponding to the rightmost time point of the segment is recorded as the second concentration. The segment slope = (second concentration - first concentration) / segment length, where the segment length is the length of the horizontal axis corresponding to the segment. The preset slope difference value can be determined by the user based on the actual application scenario. The smaller the preset slope difference value, the greater the accuracy of the data in the associated segments being at the same wear stage. A preset slope difference value is provided. The preset slope difference value is the average of the slope differences corresponding to the associated segments that meets user requirements during historical data processing.

[0096] The maximum slope deviation is confirmed by recording each associated paragraph and the spectral paragraph that is not recorded as an associated paragraph as a characteristic paragraph, detecting the characteristic slope of each characteristic paragraph, and recording the difference between the maximum characteristic slope and the minimum characteristic slope as the maximum slope deviation. The characteristic slopes of the associated paragraphs and the spectral paragraphs that are not recorded as associated paragraphs are confirmed differently. The characteristic slope of the associated paragraph is the average of the paragraph slopes of each spectral paragraph corresponding to the single associated paragraph. For a single spectral paragraph that is not recorded as an associated paragraph, the characteristic slope is the paragraph slope of the single spectral paragraph that is not recorded as an associated paragraph.

[0097] It can be understood that the values ​​of the paragraph coefficient and the slope coefficient can be obtained by users through deep learning convolutional neural network learning based on historical records. It can be understood that the present invention reflects the distribution of feature paragraphs in valid unit data through data paragraph reference values. Users can use deep learning through historical user data to obtain the number of feature paragraphs and the influence of the maximum slope deviation on the distribution of feature paragraphs in valid unit data, and then select the values ​​of the paragraph coefficient and the slope coefficient accordingly, wherein the paragraph coefficient + slope coefficient = 1, and one value of the paragraph coefficient and the slope coefficient is provided, the paragraph coefficient is 0.5, and the slope coefficient is 0.5.

[0098] The values ​​of the preset number of feature paragraphs and the preset maximum slope deviation can be determined by the user according to the actual application scenario. A preset number of feature paragraphs and the preset maximum slope deviation are provided. The preset number of feature paragraphs is the average value of the number of feature paragraphs corresponding to each valid unit data, and the preset maximum slope deviation is the average value of the maximum slope deviation corresponding to each valid unit data.

[0099] Feature threshold = data increase / preset data increase × increase coefficient + unit data duration / preset unit data duration × duration coefficient, data increase is the difference between the maximum concentration value and the minimum concentration value in a single valid unit data, unit data duration is the monitoring time length corresponding to a single valid unit data, the values ​​of the increase coefficient and the duration coefficient, users can learn from historical records through deep learning convolutional neural networks. It can be understood that the present invention reflects the influence of the effective unit data characteristics through feature thresholds, and users can use deep learning through historical user data to obtain the influence of data increase and unit data duration on the effective unit data characteristics, and then select the values ​​of the increase coefficient and duration coefficient accordingly, where the increase coefficient + duration coefficient = 1, and one value of the increase coefficient and duration coefficient is provided, the increase coefficient is 0.7, and the duration coefficient is 0.3.

[0100] The values ​​of the preset data increase and the preset unit data duration can be determined by the user according to the actual application scenario. A value of the preset data increase and the preset unit data duration is provided. The preset data increase is the average value of the data increase corresponding to each valid unit data, and the preset unit data duration is the average value of the unit data duration corresponding to each valid unit data.

[0101] The values ​​of the preset data paragraph reference value and the preset feature threshold can be determined by the user according to the actual application scenario. The higher the user's demand for the accuracy of the clutch performance degradation model prediction, the smaller the values ​​of the preset data paragraph reference value and the preset feature threshold. A value of a preset data paragraph reference value and a preset feature threshold is provided. The preset data paragraph reference value is the average value of the data paragraph reference values ​​that can meet the user's needs during the historical data processing process. The preset data paragraph reference value is the average value of the feature thresholds that can meet the user's needs during the historical data processing process.

[0102] Specifically, segmentation is performed based on concentration values;

[0103] For a valid unit data, the starting concentration point and the ending concentration point of the data segment whose concentration value is greater than the first preset concentration and less than the second preset concentration are used as segmentation points for segmentation;

[0104] The first preset concentration is less than the second preset concentration.

[0105] Among them, a data segment is a segment containing concentration values ​​corresponding to several time points. For a data segment, the starting concentration point is the concentration value corresponding to the first time point of the data segment in the time sequence, and the ending concentration point is the concentration value corresponding to the last time point of the data segment in the time sequence.

[0106] The first preset concentration is greater than 0. The values ​​of the first preset concentration and the second preset concentration can be determined by the user according to the actual application scenario. The values ​​of the first preset concentration and the second preset concentration are provided. For a valid unit data, the valid unit data records with the same elements as the valid unit data are detected during the historical data processing process. The average value of the concentration values ​​corresponding to the starting concentration points that can meet the user's segmentation requirements is recorded as the first preset concentration, and the average value of the concentration values ​​corresponding to the ending concentration points that can meet the user's segmentation requirements is recorded as the second preset concentration.

[0107] Specifically, segmentation is performed based on torque threshold;

[0108] For torque data corresponding to a valid unit of data, a torque segment having a torque reference coefficient greater than or equal to a preset torque reference coefficient is recorded as a characteristic torque segment, and distribution detection is performed on each characteristic torque segment. When distribution detection is performed on a single characteristic torque segment, the characteristic torque segment is recorded as a target characteristic torque segment. A distance reference value between each characteristic torque segment after the time sequence of the target characteristic torque segment is detected. If the distance reference value between each characteristic torque segment after the time sequence of the target characteristic torque segment is less than the preset distance reference value, the segment is recorded as an associated characteristic torque segment of the target characteristic torque segment. If the distance reference value between each characteristic torque segment after the time sequence of the target characteristic torque segment is greater than or equal to the preset distance reference value, distribution detection for the target characteristic torque segment is stopped, and the segment between the starting torque point of the target characteristic torque segment and the ending torque point of the associated characteristic torque segment with the largest distance reference value corresponding to the target characteristic torque segment is recorded as a segment combination.

[0109] The valid unit data is segmented using the starting torque point and the ending torque point corresponding to the paragraph combination with the largest horizontal coordinate length as segmentation points;

[0110] The torque threshold is determined according to a torque reference coefficient and a torque fluctuation coefficient. The torque reference coefficient is positively correlated with the torque mean value, and the torque fluctuation coefficient is positively correlated with the torque fluctuation value.

[0111] The method for confirming the torque segment is to draw a torque image corresponding to the valid unit data with the monitoring time as the horizontal axis and the torque value as the vertical axis, and use each turning point as an isolation point to obtain several torque segments;

[0112] Torque threshold = torque reference coefficient + torque fluctuation coefficient. The torque mean is determined as follows: for a single torque segment, the torque mean is the average of the torque reference values ​​at each time point corresponding to the torque segment. For a single time point, the torque reference value = the torque value at that single time point / the piston pressure value at that single time point. The piston pressure value is the net pressure applied to the clutch piston. The torque fluctuation value = maximum torque value - minimum torque value. The maximum torque value is the maximum torque of the clutch in the torque data corresponding to the valid unit data, and the minimum torque value is the minimum torque of the clutch in the torque data corresponding to the valid unit data.

[0113] For a torque segment, the first time point corresponding to the torque segment in the time direction is recorded as the starting torque point, and the last time point corresponding to the horizontal coordinate of the torque segment in the time direction is recorded as the ending torque point. The distance reference value is determined by, for two torque segments, the distance between the ending point of the first torque segment and the starting point of the second torque segment in the time direction.

[0114] The values ​​of the preset distance reference value and the preset torque fluctuation value can be determined by the user according to the actual application scenario. It can be understood that the smaller the values ​​of the preset distance reference value and the preset torque fluctuation value, the greater the accuracy of the data in the segmented section for the same wear stage, and thus the higher the accuracy of the clutch performance degradation model prediction. A preset distance reference value and a preset torque fluctuation value are provided, and the preset distance reference value is 3h. The method for confirming the preset torque fluctuation value is to detect the torque fluctuation value corresponding to the record with the same element as the valid unit data corresponding to the current target spectral data in the historical data processing process, and record the average value of the torque fluctuation value corresponding to the historical data processing process in which the segmentation effect meets the user's needs as the preset torque fluctuation value.

[0115] Specifically, whether the segmented paragraph is qualified is determined based on the comprehensive deviation value;

[0116] Detecting a starting temperature point and an ending temperature point of a temperature segment where the temperature reference value is greater than or equal to a preset temperature reference value, and determining a comprehensive deviation value based on the sum of the absolute value of the difference between the starting temperature point and the starting torque point and the absolute value of the difference between the ending temperature point and the ending torque point;

[0117] If the comprehensive deviation value is greater than or equal to the preset comprehensive deviation value, the segmentation fails, and the starting midpoint and the ending midpoint are selected as the segmentation points for segmentation;

[0118] If the comprehensive deviation value is less than the preset comprehensive deviation value, the segmentation is qualified.

[0119] The temperature reference value is the temperature of the clutch at each time point. The temperature section is a section containing the temperatures corresponding to several time points. The starting temperature point of the temperature section is the first time point corresponding to the temperature section in the time direction, and the ending temperature point is the last time point corresponding to the temperature section in the time direction.

[0120] For valid unit data, the user can determine the value of the preset temperature reference value based on the actual application scenario. A preset temperature reference value is provided. The preset temperature value is the temperature at which the clutch lasts longest during the detection process corresponding to the valid unit data. For valid unit data, the user can determine the value of the preset comprehensive deviation value based on the actual application scenario. The higher the user's demand for the accuracy of the clutch performance degradation model prediction, the smaller the preset comprehensive deviation value. A preset comprehensive deviation value is provided, and the preset comprehensive deviation value is 4 hours.

[0121] The starting midpoint = (starting temperature point + starting torque point) / 2, and the ending midpoint = (ending temperature point + ending torque point) / 2.

[0122] Specifically, the element characterization value is determined according to the sum of the unit characterization values ​​corresponding to each data point.

[0123] Among them, for a data point, the unit representation value corresponding to the data point = -(data point probability value × logarithm of the data point probability value with base 2), and the data representation value is equal to the sum of the unit representation values ​​corresponding to each data point.

[0124] The data points are the concentration values ​​corresponding to each time point, and the probability value of the data point = the number of time points corresponding to the same concentration value / the total number of time points.

[0125] In the present invention, in the first preset unit data state, the starting concentration point of the data segment whose concentration value is greater than the first preset concentration and less than the second preset concentration is recorded as the first segmentation point, and the ending concentration point of the data segment whose concentration value is greater than the first preset concentration and less than the second preset concentration is recorded as the second segmentation point;

[0126] In the second preset unit data state, if the segmentation is qualified, the starting torque point corresponding to the segment combination with the largest horizontal coordinate length is recorded as the first segmentation point, and the ending torque point corresponding to the segment combination with the largest horizontal coordinate length is recorded as the second segmentation point;

[0127] If the segmentation fails, the starting midpoint is recorded as the first segmentation point, and the ending midpoint is recorded as the second segmentation point;

[0128] The concentration values ​​corresponding to the first segmentation point and the concentration values ​​corresponding to each time point before the first segmentation point in the time sequence are recorded as the first segmentation section, the concentration values ​​corresponding to the time points between the first segmentation point and the second segmentation point are recorded as the second segmentation section, and the concentration values ​​corresponding to the second segmentation point and the concentration values ​​corresponding to each time point after the second segmentation point in the time sequence are recorded as the third segmentation section.

[0129] Among them, the wear stage corresponding to the first segmented paragraph is the initial wear stage, the wear stage corresponding to the second segmented paragraph is the stable wear stage, and the wear stage corresponding to the third segmented paragraph is the severe wear stage. The first segmented paragraph, the second segmented paragraph, and the third segmented paragraph corresponding to each valid unit data are modeled and predicted using the Wiener process respectively. This is content that is easy for technical personnel in this field to understand and will not be elaborated on in detail.

[0130] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

[0131] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A clutch intelligent detection method based on multi-dimensional data acquisition, characterized in that: include: Get clutch related data uploaded by users; determining target spectral data according to the vehicle characterization coefficient; Determine the valid unit data according to the element characterization value, and determine the target spectrum data according to the number of valid unit data. The data screening method is to screen according to the correlation coefficient, or to screen Ф valid unit data in descending order according to the element characterization value; Determine the unit data state according to the data segment reference value and the feature threshold; Determining, according to the unit data status of the valid unit data, a data processing method of the valid unit data, that is, segmenting according to a concentration value, or segmenting according to a torque threshold; Under the preset conditions, whether the segmented paragraph is qualified is determined based on the comprehensive deviation value; The preset condition is that the segmentation is completed according to the torque threshold; The element characterization value is determined based on the sum of the unit characterization values ​​corresponding to each data point, where the unit characterization value corresponding to the data point = -(data point probability value × logarithm of the data point probability value with base 2), and the data point probability value = the number of time points corresponding to the same concentration value / the total number of time points; The vehicle characterization coefficient is determined according to the block frequency reference value and the environment characterization value; The relationship between the vehicle characterization coefficient and the frequency blocking reference value is a positive correlation relationship, and the relationship between the vehicle characterization coefficient and the environment characterization value is a positive correlation relationship; The comprehensive deviation value is determined according to the sum of the absolute value of the difference between the starting temperature point and the starting torque point and the absolute value of the difference between the ending temperature point and the ending torque point.

2. The intelligent clutch detection method based on multi-dimensional data acquisition according to claim 1 is characterized in that: determining target spectral data according to the vehicle characterization coefficient; Dividing the data sequence of the spectral data and obtaining a number of pre-screened spectral data, each pre-screened spectral data being recorded as target spectral data; The data sequence of the spectral data is the order of vehicle characterization coefficients from large to small.

3. The intelligent clutch detection method based on multi-dimensional data acquisition according to claim 2 is characterized in that: Determine valid unit data according to the element characterization value, and determine the data screening method of the target spectrum data according to the number of valid unit data; If the number of valid unit data is greater than or equal to the preset number, the data screening method is to filter according to the correlation coefficient; If the number of valid unit data is less than the preset number, the data screening method is to screen Ф valid unit data in descending order according to the element representation value; The valid unit data is unit data whose element representation value is greater than a preset element representation value.

4. The intelligent clutch detection method based on multi-dimensional data acquisition according to claim 3 is characterized in that: Filtering based on correlation coefficients includes: For a target spectral data, the valid unit data with the largest element characterization value in the target spectral data is recorded as the target unit, and the other valid unit data in the target spectral data excluding the target unit is recorded as the reference unit. The correlation coefficient between the target unit and each reference unit is detected, and Φ-1 reference units are screened in descending order of the correlation coefficient. The target unit and the screened Φ-1 reference units are recorded as screened data.

5. The intelligent clutch detection method based on multi-dimensional data acquisition according to claim 4 is characterized in that: Determine the data processing mode of the valid unit data according to the unit data status of the valid unit data; If the unit data state is the first preset unit data state, segmentation is performed according to the concentration value; If the unit data state is the second preset unit data state, segmentation is performed according to the torque threshold.

6. The intelligent clutch detection method based on multi-dimensional data acquisition according to claim 5 is characterized in that: The unit data state is determined based on the data segment reference value and the characteristic threshold value, and the unit data state includes: A first preset unit data state in which the data segment reference value is less than a preset data segment reference value and the characteristic threshold is less than a preset characteristic threshold; A second preset unit data state in which the data segment reference value is greater than or equal to the preset data segment reference value or the feature threshold is greater than or equal to the preset feature threshold; The data segment reference value is determined according to the number of characteristic segments and the maximum slope deviation, and the characteristic threshold is determined according to the data increase and the unit data duration.

7. The intelligent clutch detection method based on multi-dimensional data acquisition according to claim 5 is characterized in that: Segmentation based on concentration values; For a valid unit data, the starting concentration point and the ending concentration point of the data segment whose concentration value is greater than the first preset concentration and less than the second preset concentration are used as segmentation points for segmentation; The first preset concentration is less than the second preset concentration.

8. The intelligent clutch detection method based on multi-dimensional data acquisition according to claim 5 is characterized in that: Segmentation based on torque threshold; For torque data corresponding to a valid unit of data, a torque segment having a torque reference coefficient greater than or equal to a preset torque reference coefficient is recorded as a characteristic torque segment, and distribution detection is performed on each characteristic torque segment. When distribution detection is performed on a single characteristic torque segment, the characteristic torque segment is recorded as a target characteristic torque segment. A distance reference value between each characteristic torque segment after the time sequence of the target characteristic torque segment is detected. If the distance reference value between each characteristic torque segment after the time sequence of the target characteristic torque segment is less than the preset distance reference value, the segment is recorded as an associated characteristic torque segment of the target characteristic torque segment. If the distance reference value between each characteristic torque segment after the time sequence of the target characteristic torque segment is greater than or equal to the preset distance reference value, distribution detection for the target characteristic torque segment is stopped, and the segment between the starting torque point of the target characteristic torque segment and the ending torque point of the associated characteristic torque segment with the largest distance reference value corresponding to the target characteristic torque segment is recorded as a segment combination. The valid unit data is segmented using the starting torque point and the ending torque point corresponding to the paragraph combination with the largest horizontal coordinate length as segmentation points; The torque threshold is determined according to a torque reference coefficient and a torque fluctuation coefficient. The torque reference coefficient is positively correlated with the torque mean value, and the torque fluctuation coefficient is positively correlated with the torque fluctuation value.

9. The intelligent clutch detection method based on multi-dimensional data acquisition according to claim 5, characterized in that: Determine whether the segmented paragraph is qualified according to the comprehensive deviation value; Detecting the starting temperature point and the ending temperature point of the temperature segment whose temperature reference value is greater than or equal to the preset temperature reference value; If the comprehensive deviation value is greater than or equal to the preset comprehensive deviation value, the segmentation fails, and the starting midpoint and the ending midpoint are selected as the segmentation points for segmentation; If the comprehensive deviation value is less than the preset comprehensive deviation value, the segmentation is qualified.

Images