Intelligent monitoring system of transport posture of three-dimensional impact recorder based on sensor technology

Abstract

The present invention relates to the field of data processing technology, and specifically to a three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology, which collects three-dimensional data and collection time of impact data; obtains several clusters and a three-dimensional data point sequence of each cluster; obtains the driving stage characteristics of the impact data of each cluster in each dimension; obtains the impact data heterogeneity of each data point in each dimension of the three-dimensional data point sequence of each cluster, and then obtains the specific factor of each data point in the three-dimensional data point sequence of each cluster; obtains the distance between data points; obtains the optimal k-neighbor distance, obtains the outliers in the impact data through the LOF algorithm; processes the outliers in the impact data, and monitors the posture changes through the processed impact data. The present invention improves the accuracy of transportation posture analysis by reducing the influence of data changes caused by different transportation stages on the outlier detection process.

Description

Technical Field

[0001] The invention relates to the technical field of data processing, and in particular to a three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology. Background Art

[0002] A 3D shock recorder, also known as a transport vibrometer or impact data recorder, is a device used to monitor and record the shock and vibration conditions of an object during transportation. It is usually equipped with a three-axis acceleration sensor that can sense and record shock and vibration data in three directions (X, Y, and Z axes), and these data are mainly stored in the form of acceleration data in the three-axis directions.

[0003] In the process of using a three-dimensional impact recorder to monitor the transportation posture, the three-axis acceleration direction data collected by the recorder is usually used for analysis and representation. However, in the actual transportation process, the acceleration sensor in the recorder may be disturbed by the transportation environment when collecting data. These interferences may be caused by electronic interference from surrounding equipment during transportation or interference from transportation equipment. In order to accurately analyze and monitor the transportation posture, it is very important to ensure the accuracy of the impact data collected by the recorder. Summary of the invention

[0004] The invention provides a three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology to solve the existing problems.

[0005] The three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology of the present invention adopts the following technical solutions:

[0006] The present invention proposes a three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology, which includes the following modules:

[0007] Data acquisition module: used to collect three-dimensional data and acquisition time of impact data;

[0008] Data cleaning module: used to cluster the three-dimensional data and acquisition time of the impact data to obtain several clusters and a three-dimensional data point sequence for each cluster;

[0009] According to the difference between the accelerations of different data points in the same dimension in the three-dimensional data point sequence of each cluster, the driving stage characteristics of the impact data of each cluster in each dimension are obtained; combined with the acceleration changes of adjacent data points in the three-dimensional data point sequence of each cluster in the same dimension, the impact data heterogeneity of each data point in each cluster's three-dimensional data point sequence is obtained, and then the specific factor of each data point in the three-dimensional data point sequence of each cluster is obtained;

[0010] According to the specific factors of each data point in the three-dimensional data point sequence of each cluster and the acceleration value in each dimension, as well as the driving stage characteristics of the impact data of each cluster in each dimension, combined with the collection time of the impact data, the distance between the data points is obtained; according to the specific factors of each data point, combined with the distance between the data points, the evaluation function of the k-neighbor distance is constructed, and then the optimal k-neighbor distance is obtained, and the outliers in the impact data are obtained through the LOF algorithm;

[0011] Data processing module: used to process abnormal values ​​in impact data and monitor posture changes through the processed impact data.

[0012] Furthermore, the three-dimensional data and acquisition time based on the impact data are clustered to obtain a number of clusters and a three-dimensional data point sequence of each cluster, including the specific method of:

[0013] The collected three-dimensional data is combined with the collection time to obtain four-dimensional data about the impact data;

[0014] The four-dimensional data obtained by scanning are clustered by K-means clustering algorithm. The distance measurement adopts the Euclidean distance between the four-dimensional data to divide the four-dimensional data into K clusters.

[0015] The acquisition time dimension of the four-dimensional data in each cluster is removed to obtain K clusters consisting of three-dimensional data, and the data points corresponding to the three-dimensional data in the same cluster are arranged in the order of acquisition time to obtain a three-dimensional data point sequence for each cluster.

[0016] Furthermore, according to the difference between the accelerations of different data points in the three-dimensional data point sequence of each cluster in the same dimension, the driving phase characteristics of the impact data of each cluster in each dimension are obtained, and the corresponding specific calculation formula is:

[0017]

[0018] Among them, C u,v N represents the driving phase feature value of the u-th cluster in the v-th dimension impact data; u A represents the number of data points in the three-dimensional data point sequence of the u-th cluster, u,n,v A represents the acceleration value of the vth dimension of the nth data point in the three-dimensional data point sequence of the uth cluster; u,n±1,v represents the acceleration value of the vth dimension of the n+1th data point and the acceleration value of the vth dimension of the n-1th data point in the three-dimensional data point sequence of the uth cluster; || represents the absolute value function; It represents the mean of the absolute values ​​of the differences between the acceleration values ​​of the nth data point and the two adjacent data points on the left and right in the three-dimensional data point sequence of the uth cluster in the vth dimension; norm represents the normalization function.

[0019] Furthermore, the specific calculation formula corresponding to the impact data heterogeneity of each dimension of each data point in the three-dimensional data point sequence of each cluster is:

[0020]

[0021] in, A represents the impact data heterogeneity of the vth dimension of the nth data point in the three-dimensional data point sequence of the uth cluster; u,n,v A represents the acceleration value of the vth dimension of the nth data point in the three-dimensional data point sequence of the uth cluster; u,n+1,v A represents the acceleration value of the vth dimension of the n+1th data point in the three-dimensional data point sequence of the uth cluster; u,n-1,v represents the acceleration value of the vth dimension of the n-1th data point in the three-dimensional data point sequence of the uth cluster; C u,v represents the driving stage eigenvalue of the u-th cluster in the v-th dimension impact data; || represents the absolute value function; tanh() represents the hyperbolic tangent function.

[0022] Furthermore, the specific factor of each data point in the three-dimensional data point sequence of each cluster is obtained, and the corresponding specific calculation formula is:

[0023]

[0024] Among them, γ u,n Represents the specific factor of the nth data point in the three-dimensional data point sequence of the uth cluster; Represents the impact data heterogeneity of the x-axis dimension of the n-th data point in the three-dimensional data point sequence of the u-th cluster; Represents the impact data heterogeneity of the y-axis dimension of the n-th data point in the three-dimensional data point sequence of the u-th cluster; It represents the impact data heterogeneity of the z-axis dimension of the nth data point in the three-dimensional data point sequence of the uth cluster; norm() represents the normalization function; E() represents the mean calculation function.

[0025] Furthermore, the specific factor of each data point in the three-dimensional data point sequence of each cluster and the acceleration value in each dimension, as well as the driving stage characteristics of the impact data of each cluster in each dimension, combined with the collection time of the impact data, the distance between the data points is obtained, and the corresponding specific calculation formula is:

[0026]

[0027] Among them, d a,b represents the distance between the a-th data point and the b-th data point; γ a represents the specific factor of the a-th data point; γ b represents the specific factor of the bth data point; ΔA a,b,x Indicates the square of the difference in acceleration between the a-th data point and the b-th data point in the dimension of the x-axis; ΔA a,b,y Indicates the square of the difference in acceleration between the a-th data point and the b-th data point in the dimension of the y-axis; ΔA a,b,z Indicates the square of the difference in acceleration between the a-th data point and the b-th data point in the dimension of the z-axis; Δt a,b Indicates the square of the time series difference between the a-th data point and the b-th data point; ΔC a,b It represents the sum of square differences between the driving stage eigenvalues ​​of the cluster where the a-th data point belongs and the cluster where the b-th data point belongs in three dimensions.

[0028] Furthermore, the evaluation function of k-neighbor distance is constructed based on the specific factor of each data point and the distance between data points. The corresponding specific calculation formula is:

[0029]

[0030] Among them, Q represents the output value of the evaluation function; I represents the number of data points; γ i represents the specific factor of the i-th data point; Represents the local density of the k-neighborhood of the i-th data point; represents the local density mean of the k-neighborhood of all data points; J i,k represents the number of data points in the k-neighborhood of the ith data point; d i,j Represents the distance between the i-th data point and the j-th data point in the k-neighborhood of the i-th data point.

[0031] Furthermore, the specific method of obtaining the optimal k-neighbor distance includes:

[0032] The k-neighbor distance when the evaluation function takes the maximum value during the iteration process is recorded as the optimal k-neighbor distance.

[0033] Furthermore, the specific method of obtaining the outliers in the impact data by using the LOF algorithm is as follows:

[0034] The optimal k-neighbor distance is used as the parameter of the LOF algorithm to detect outliers in the impact data and obtain the outliers in the impact data.

[0035] Furthermore, the specific method of processing the abnormal values ​​in the impact data includes:

[0036] The impact data are sorted in time series, and outliers are deleted. The deleted impact data values ​​are filled in by data interpolation to obtain the processed impact data.

[0037] The beneficial effects of the technical solution of the present invention are:

[0038] Through data clustering, specific factor analysis and outlier detection, the accuracy of posture monitoring during transportation is improved, the impact of different transportation stages on data changes is reduced, the accuracy of outlier detection is improved, and finally the intelligent monitoring of transportation posture is realized. Specifically, the K-means algorithm is used for clustering based on the collected three-dimensional data and the collection time. The optimal clustering number K is obtained by the elbow method, and the accuracy of the clustering results is ensured by using the Euclidean distance measurement. The driving stage characteristic values ​​of each cluster in each dimension are calculated to clarify the characteristic differences of each stage. By comparing the acceleration differences of data points in the same cluster, the specific impact characteristics of each stage are revealed. The accuracy of the specific factors is ensured by the acceleration values ​​of each data point in different dimensions and the differences with adjacent data points. The distance measurement method between data points is improved based on the specific factors and acceleration values. Through the normalization and mean calculation functions, a more accurate distance evaluation function is constructed to improve the measurement accuracy of data relationships. The impact data is detected using the LOF algorithm and the optimal k-neighbor distance parameter to ensure the accuracy and reliability of outlier identification. Finally, the detected outliers are cleaned, deleted and filled with data interpolation to ensure the continuity and integrity of the processed data, thereby improving the accuracy of posture monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 This is a module flow chart of the three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology of the present invention. DETAILED DESCRIPTION

[0041] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation method, structure, features and effects of the three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology proposed by the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0042] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0043] The specific scheme of the three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology provided by the present invention is described in detail below with reference to the accompanying drawings.

[0044] See also Figure 1 , which shows a module flow chart of a three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology provided by an embodiment of the present invention, the system includes the following modules:

[0045] Module 101: Data acquisition module, which collects three-dimensional data and acquisition time of impact data.

[0046] The purpose of this embodiment is to analyze the three-dimensional data recorded by the three-dimensional impact recorder, design a data cleaning method, and analyze and monitor the transportation posture. Therefore, it is necessary to collect the three-dimensional acceleration data and the corresponding collection time in the impact data through the three-dimensional impact recorder during the process of monitoring the transportation posture.

[0047] It should be noted that in the process of monitoring the transportation posture, it is necessary to collect three-dimensional acceleration data through a three-dimensional impact recorder. The operating principle of the three-dimensional impact recorder is to detect the impact and vibration of the object in all directions during transportation through the built-in three-axis acceleration sensor. When the device senses impact or vibration, the acceleration sensor will generate a corresponding electrical signal, which will be converted into readable data by the recorder. The device can collect data continuously and is powered by a disposable lithium battery to ensure long-term monitoring and recording. The recorded data includes three-axis acceleration values, temperature and humidity, etc.

[0048] Specifically, a suitable recorder is selected according to the size of the object to be monitored, the mode of transportation, and the environmental conditions; the recorder is installed on the object to ensure that its position is stable and can accurately record the movement and posture changes of the object; the impact recorder is started to start recording the impact and vibration of the object during transportation; the three-axis acceleration values ​​and their corresponding collection times are obtained, and the three-axis acceleration values ​​are combined into three-dimensional data, in which the different dimensions of the data are the three-axis accelerations.

[0049] At this point, the impact data has been collected.

[0050] Module 102: A data cleaning module, which performs clustering based on the three-dimensional data and acquisition time of the impact data to obtain a number of clusters and a three-dimensional data point sequence for each cluster.

[0051] It should be noted that when obtaining the impact data of the transportation process, the data collection and transmission stages may be disturbed by the transportation process, so the collected impact data needs to be analyzed. However, since the acceleration directions of the transportation equipment are different in different stages of the transportation process, direct analysis of the data may be disturbed by the normal changes in the acceleration of the transportation process, reducing the accuracy of the analysis results. Therefore, it is necessary to clean the collected data.

[0052] Specifically, the collected three-dimensional data is combined with the collection time to obtain four-dimensional data about the impact data; the scanned four-dimensional data is clustered by the K-means clustering algorithm, wherein the K value is obtained by the elbow method, and the distance measurement adopts the Euclidean distance between the four-dimensional data to divide the four-dimensional data into K clusters. It should be noted that the K-means clustering algorithm and the elbow method are common and well-known algorithms, and the specific method will not be repeated here; the collection time dimension of the four-dimensional data in each cluster is removed to obtain K clusters composed of three-dimensional data, and the data points corresponding to the three-dimensional data in the same cluster are arranged in the order of collection time to obtain a three-dimensional data point sequence for each cluster.

[0053] It should be noted that after the data is divided into several clusters through the clustering algorithm, it is necessary to analyze the outliers in the data, by combining the data change characteristics of the transportation stage reflected in different clusters and the characteristics of each data point in the cluster, and adjusting the three-dimensional data anomaly analysis process.

[0054] (1) Based on the differences in acceleration of different data points in the three-dimensional data point sequence of each cluster in the same dimension, the driving stage characteristics of the impact data of each cluster in each dimension are obtained; combined with the acceleration changes of adjacent data points in the three-dimensional data point sequence of each cluster in the same dimension, the impact data heterogeneity of each data point in the three-dimensional data point sequence of each cluster is obtained, and then the specific factor of each data point in the three-dimensional data point sequence of each cluster is obtained.

[0055] It should be noted that the impact data of different clusters mainly reflect the data change characteristics of different stages in the transportation process. During the transportation process, there may be driving stages such as sudden braking and sharp turns, which may cause the impact data to suddenly increase, and its manifestation is similar to abnormal data. In order to reduce the interference of normal data changes on the abnormal detection process, it is necessary to judge the abnormal information of the data point before detecting the abnormal data. The abnormal information of the data point is related to the driving stage. Therefore, it is necessary to first analyze the driving stage characteristics of different data point clusters.

[0056] Specifically, the specific calculation formula for the driving stage characteristics of each cluster in each dimension of impact data is:

[0057]

[0058] Among them, C u,v N represents the driving phase feature value of the u-th cluster in the v-th dimension impact data; u A represents the number of data points in the three-dimensional data point sequence of the u-th cluster, u,n,v A represents the acceleration value of the vth dimension of the nth data point in the three-dimensional data point sequence of the uth cluster; u,n±1,v represents the acceleration value of the vth dimension of the n+1th data point and the acceleration value of the vth dimension of the n-1th data point in the three-dimensional data point sequence of the uth cluster; || represents the absolute value function; represents the mean of the absolute values ​​of the differences between the acceleration values ​​of the nth data point and the two adjacent data points on the left and right in the three-dimensional data point sequence of the uth cluster; norm represents the linear normalization function, and the normalization object is all the data points in the three-dimensional data point sequence of each cluster.

[0059] It should be noted that after clustering the three-dimensional data of the impact recorder, different clusters represent different driving stages. represents the acceleration mean of the vth dimension of each data point in the uth cluster. In order to reduce the interference of outliers in three-dimensional data on the overall form stage characteristics, it is also necessary to perform weighted processing on the mean calculation process. Calculate the mean of the absolute value of the difference between the acceleration values ​​of the nth data point and its adjacent data points in the vth dimension in the three-dimensional data point sequence of the uth cluster; in the general driving stage, the acceleration change to the impact recorder is continuous and gentle. When the mean is large, the probability that the acceleration value of the nth data point is an abnormal value is large, so through After the difference is standardized, the corrected logical relationship is used as the mean calculation coefficient, replacing The weights of different data points in the mean calculation process are adjusted to obtain the driving stage feature value of the u-th cluster impacting the data in the v-th dimension.

[0060] It should be noted that after obtaining the driving stage characteristics of each cluster in each dimension of the impact data, to analyze the specific factors of a certain data point, it is necessary to analyze the difference between the data point and other data points in the cluster to which it belongs. The cluster to which it belongs is usually a certain driving stage in the transportation process. The impact data in this stage generally changes continuously and smoothly in time series; for abnormal data points, there is a large difference in the acceleration value of this data point and other data points in the stage to which it belongs. Therefore, the abnormal information of the data point can be analyzed by combining the changes in the acceleration values ​​of other data points in the cluster to which the data point belongs.

[0061] Specifically, the specific calculation formula for the specific factor of each data point in the three-dimensional data point sequence of each cluster is:

[0062]

[0063] in, A represents the impact data heterogeneity of the vth dimension of the nth data point in the three-dimensional data point sequence of the uth cluster; u,n,v A represents the acceleration value of the vth dimension of the nth data point in the three-dimensional data point sequence of the uth cluster; u,n+1,v A represents the acceleration value of the vth dimension of the n+1th data point in the three-dimensional data point sequence of the uth cluster; u,n-1,v represents the acceleration value of the vth dimension of the n-1th data point in the three-dimensional data point sequence of the uth cluster; C u,v represents the driving phase eigenvalue of the u-th cluster impacting the data in the v-th dimension; γ u,n Represents the specific factor of the nth data point in the three-dimensional data point sequence of the uth cluster; Represents the impact data heterogeneity of the x-axis dimension of the n-th data point in the three-dimensional data point sequence of the u-th cluster; Represents the impact data heterogeneity of the y-axis dimension of the n-th data point in the three-dimensional data point sequence of the u-th cluster; represents the impact data heterogeneity of the z-axis dimension of the nth data point in the three-dimensional data point sequence of the uth cluster; || represents the absolute value function; tanh() represents the hyperbolic tangent function; norm() represents the linear normalization function, and the normalization object is all data points in all clusters. E() represents the mean calculation function.

[0064] It should be noted that, since the data points with large differences in characteristics from the current driving stage are more likely to be abnormal data, the specific factors of the data points can be analyzed by comparing the data points in the local area with the current data points. u,n,v -C u,v|The difference between the acceleration value of the vth dimension of the nth data point in the three-dimensional data point sequence of the uth cluster and the driving stage characteristic value of the impact data of the uth cluster in the vth dimension is calculated. The larger the difference, the greater the abnormality of the nth data point in the current cluster. Secondly, the data in the above-mentioned special driving stage also has certain abnormal characteristics. In order to reduce the interference of these special driving stage data on the abnormal calculation process, it is necessary to adjust the difference calculation results; because in the above-mentioned special driving stage, the acceleration suddenly increases at the previous time point and gradually decreases at the subsequent time points; and the changes between the abnormal data points and the adjacent data points are large. Therefore, the specific factors of the data can be analyzed by the rate of change between adjacent data points; The change rates between the nth data point and the previous and next data points are calculated respectively. Under normal circumstances, these two change rates are usually smaller. In some special transportation stages, one change rate may be larger and the other change rate may be smaller. However, both change rates of abnormal values ​​are larger. After adding them and normalizing the value range, we can get the value of As an adjustment factor, with |A u,n,v -C u,v | are multiplied, and the product is used as the impact data heterogeneity of the vth dimension of the nth data point in the three-dimensional data point sequence of the uth cluster; after averaging the impact data heterogeneity of the x, y, and z axes, the specific factor of the nth data point in the three-dimensional data point sequence of the uth cluster is obtained.

[0065] (2) According to the specific factors of each data point in the three-dimensional data point sequence of each cluster and the acceleration value in each dimension, as well as the driving stage characteristics of the impact data of each cluster in each dimension, combined with the collection time of the impact data, the distance between the data points is obtained; according to the specific factors of each data point and the distance between the data points, an evaluation function of the k-neighbor distance is constructed, and then the optimal k-neighbor distance is obtained, and the outliers in the impact data are obtained through the LOF algorithm.

[0066] It should be noted that after obtaining the specific factors of each data point in the three-dimensional data point sequence of each cluster, it is necessary to further analyze the abnormalities in the data space. The outliers of the data points can be represented by the degree of outliers of the data points. Therefore, it is necessary to optimize the distance calculation process between the data points in combination with the needs of outlier analysis. The data point distance can be calculated by analyzing the time series distance of the data points and the difference in the impact data in combination with the specific factors of the data points.

[0067] Specifically, for all data points in the three-dimensional data point sequence of all clusters, the specific calculation formula for the distance between data points is as follows:

[0068]

[0069] Among them, da,b represents the distance between the a-th data point and the b-th data point; γ a represents the specific factor of the a-th data point; γ b represents the specific factor of the bth data point; ΔA a,b,x Indicates the square of the difference in acceleration between the a-th data point and the b-th data point in the dimension of the x-axis; ΔA a,b,y Indicates the square of the difference in acceleration between the a-th data point and the b-th data point in the dimension of the y-axis; ΔA a,b,z Indicates the square of the difference in acceleration between the a-th data point and the b-th data point in the dimension of the z-axis; Δt a,b Indicates the square of the time series difference between the a-th data point and the b-th data point; ΔC a,b It represents the sum of square differences between the driving stage eigenvalues ​​of the cluster where the a-th data point belongs and the cluster where the b-th data point belongs in three dimensions.

[0070] It should be noted that, since the impact data characteristics at different stages of transportation are different, when analyzing the outliers of data points, it is necessary to minimize the impact caused by the different impact data characteristics at different stages. The mean of the specific factors of the a-th data point and the b-th data point was calculated. For data points with a relatively small overall abnormality, direct calculation through three-dimensional data is easily affected by the difference in the characteristics of the impact data in different transportation stages. When calculating their distance, the difference in the characteristics of the driving stage (Δt a,b +ΔC a,b ). For data points with a large overall abnormality, their abnormal information is mainly reflected by the impact data value, so Enlarge the calculation proportion of the three-dimensional impact data value.

[0071] It should be noted that in the data space, the abnormality of the data point can be reflected by the distance between the data points, and the outlier is represented by the local density difference in the data space. The outlier detection process can be optimized by obtaining the LOF algorithm parameters, and the optimal k-neighbor distance can be obtained by establishing an evaluation function to evaluate different k-neighbor distances, thereby obtaining abnormal data through the LOF algorithm parameters.

[0072] Specifically, the optimal k-neighbor distance is obtained by iterating the k-neighbor distance from small to large and evaluating it. It should be noted that the iteration range of the k-neighbor distance is the maximum number of data points in each cluster; the Chinese name of the LOF algorithm is the local anomaly factor detection algorithm, which is a common and well-known algorithm, and the specific method will not be repeated. The specific expression of the evaluation function is:

[0073]

[0074] Among them, Q represents the output value of the evaluation function; I represents the number of data points; γ i represents the specific factor of the i-th data point; represents the local density of the k-neighborhood of the ith data point, where the local density is the reciprocal of the mean distance between any two data points in the k-neighborhood of the ith data point. represents the local density mean of the k-neighborhood of all data points; J i,k represents the number of data points in the k-neighborhood of the ith data point; d i,j Represents the distance between the i-th data point and the j-th data point in the k-neighborhood of the i-th data point.

[0075] It should be noted that in the LOF algorithm, outliers in the data are analyzed by analyzing the local density of data points. Therefore, the ideal k-neighbor distance can distinguish the size of the k-neighborhood between abnormal data points and normal data points; The variance of the local density of the k-neighborhood of each data point under the current iteration of the k-neighborhood distance is calculated. The larger the value, the greater the difference in the size of the k-neighborhood of the data point can be. In the variance calculation process, the specific factor γ of the data point is used. i Adjust the variance calculation process to increase the weight of data points with higher abnormality in the calculation process, so as to highlight the local density difference of the k-neighborhood of the abnormal data point. In addition to the difference between the k-neighborhoods of each data point, the distance between each data point within the k-neighborhood should be as small as possible to represent the characteristics of the same transportation stage, and it can reduce the risk of excessive k-neighborhood distance iteration. When the k-neighborhood distance is too large, it may cause the local density of normal points to have abnormal point characteristics. Therefore, The mean distance between each data point and all data points in its k-neighborhood is calculated and used as the penalty term of the evaluation function. After adding it to the weighted variance, Q is used to represent the objective function in the k-neighborhood distance iteration process.

[0076] Furthermore, the k-neighbor distance at which the evaluation function takes the maximum value during the iteration process is recorded as the optimal k-neighbor distance; the optimal k-neighbor distance is used as a parameter of the LOF algorithm to perform outlier detection on the impact data to obtain the outliers in the impact data.

[0077] Module 103: a data processing module, which processes abnormal values ​​in the impact data and monitors posture changes through the processed impact data.

[0078] Specifically, the impact data is sorted in time series and the abnormal values ​​are deleted. Since the impact data monitors the acceleration values ​​in different directions during the transportation process and has a certain continuity, the deleted impact data values ​​are filled in by data interpolation to obtain the processed impact data. Then, the processed impact data is used to calculate the posture change of the object during transportation through the Euler angle solution algorithm. It should be noted that the Euler angle solution algorithm is a common and well-known technology, and the specific algorithm is not repeated here.

[0079] At this point, this embodiment is completed.

[0080] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. The three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology is characterized by: The system includes the following modules: Data acquisition module: used to collect three-dimensional data and acquisition time of impact data; Data cleaning module: used to cluster the three-dimensional data and acquisition time of the impact data to obtain several clusters and a three-dimensional data point sequence for each cluster; According to the difference between the accelerations of different data points in the same dimension in the three-dimensional data point sequence of each cluster, the driving phase characteristics of the impact data of each cluster in each dimension are obtained; Combining the acceleration changes of adjacent data points in the three-dimensional data point sequence of each cluster in the same dimension, the impact data heterogeneity of each data point in the three-dimensional data point sequence of each cluster is obtained, and then the specific factor of each data point in the three-dimensional data point sequence of each cluster is obtained; According to the specific factors of each data point in the three-dimensional data point sequence of each cluster and the acceleration value in each dimension, as well as the driving stage characteristics of the impact data of each cluster in each dimension, combined with the collection time of the impact data, the distance between the data points is obtained; According to the specific factors of each data point in the three-dimensional data point sequence of each cluster and the acceleration value in each dimension, as well as the driving stage characteristics of the impact data of each cluster in each dimension, combined with the collection time of the impact data, the distance between the data points is obtained, and the corresponding specific calculation formula is: ； in, Indicates data points and The distance between data points; Indicates Specific factors for each data point; Indicates Specific factors for each data point; Indicates data points and Data points in The square of the difference in acceleration values ​​in the dimension where the axis is located; Indicates data points and Data points in The square of the difference in acceleration values ​​in the dimension where the axis is located; Indicates data points and Data points in The square of the difference in acceleration values ​​in the dimension where the axis is located; Indicates data points and The square of the time series difference of data points; Indicates The cluster to which the data point belongs and The sum of square differences of the driving stage eigenvalues ​​of the clusters to which the data points belong in the three dimensions; According to the specific factors of each data point, the evaluation function of k-neighbor distance is constructed in combination with the distance between data points, and then the optimal k-neighbor distance is obtained, and the outliers in the impact data are obtained through the LOF algorithm; Data processing module: used to process abnormal values ​​in impact data and monitor posture changes through the processed impact data.

2. According to claim 1, the three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology is characterized in that: The three-dimensional data and acquisition time based on the impact data are clustered to obtain a number of clusters and a three-dimensional data point sequence for each cluster, including the following specific methods: The collected three-dimensional data is combined with the collection time to obtain four-dimensional data about the impact data; The four-dimensional data obtained by scanning are clustered by K-means clustering algorithm. The distance measurement adopts the Euclidean distance between the four-dimensional data to divide the four-dimensional data into K clusters. The acquisition time dimension of the four-dimensional data in each cluster is removed to obtain K clusters consisting of three-dimensional data, and the data points corresponding to the three-dimensional data in the same cluster are arranged in the order of acquisition time to obtain a three-dimensional data point sequence for each cluster.

3. According to claim 1, the three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology is characterized in that: According to the difference between the accelerations of different data points in the same dimension in the three-dimensional data point sequence of each cluster, the driving phase characteristics of the impact data of each cluster in each dimension are obtained, and the corresponding specific calculation formula is: ； in, Indicates The cluster in The driving phase characteristic values ​​of the impact data of each dimension; Indicates The number of data points in the three-dimensional data point sequence of clusters, Indicates The first The data point The acceleration value of each dimension; Indicates The first The data point The acceleration value of the dimension and the The data point The acceleration value of each dimension; represents the absolute value function; Indicates The first The first data point and the second data point adjacent to it The mean of the absolute values ​​of the differences between the acceleration values ​​in the dimensions; Represents the normalization function.

4. According to claim 1, the three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology is characterized in that: The specific calculation formula corresponding to the impact data heterogeneity of each dimension of each data point in the three-dimensional data point sequence of each cluster is: ； in, Indicates The first The data point Shock data heterogeneity in 2 dimensions; Indicates The first The data point The acceleration value of each dimension; Indicates The first The data point The acceleration value of each dimension; Indicates The first The data point The acceleration value of each dimension; Indicates The cluster in The driving phase characteristic values ​​of the impact data of each dimension; represents the absolute value function; represents the hyperbolic tangent function.

5. According to claim 1, the three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology is characterized in that: The specific calculation formula corresponding to the specific factor of each data point in the three-dimensional data point sequence of each cluster is: ； in, Indicates The first Specific factors for each data point; Indicates The first data points Heterogeneity of shock data along the dimension where the axis is located; Indicates The first data points Heterogeneity of shock data along the dimension where the axis is located; Indicates The first data points Heterogeneity of shock data along the dimension where the axis is located; represents the normalization function; Represents the mean calculation function.

6. According to claim 1, the three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology is characterized in that: The evaluation function of k-neighbor distance is constructed based on the specific factor of each data point and the distance between data points. The corresponding specific calculation formula is: ； in, Represents the output value of the evaluation function; Indicates the number of data points; Indicates Specific factors for each data point; Represents the local density of the k-neighborhood of the i-th data point; Represents the local density mean of the k-neighborhood of all data points; Indicates The number of data points in the k-neighborhood of a data point; Indicates data points and The k-neighborhood of a data point The distance between data points.

7. According to claim 1, the three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology is characterized in that: The specific method of obtaining the optimal k-neighbor distance includes: The k-neighbor distance when the evaluation function takes the maximum value during the iteration process is recorded as the optimal k-neighbor distance.

8. According to claim 1, the three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology is characterized in that: The specific method of obtaining the outliers in the impact data by using the LOF algorithm is as follows: The optimal k-neighbor distance is used as the parameter of the LOF algorithm to detect outliers in the impact data and obtain the outliers in the impact data.

9. The three-dimensional impact recorder transportation posture intelligent monitoring system based on sensor technology according to claim 1 is characterized in that: The specific method of processing the abnormal values ​​in the impact data includes: The impact data are sorted in time series, and outliers are deleted. The deleted impact data values ​​are filled in by data interpolation to obtain the processed impact data.

Images