A forestry waste crushing device and method

Abstract

The present invention relates to the technical field of clustering data processing, and particularly relates to a forestry waste crushing device and method. Multiple vibration data periods and power data periods of a crusher are obtained and clustered to obtain multiple clusters; abnormal periods are screened out based on the probability of failure occurring in each vibration data period in the cluster. Based on the probability of failure occurring in each abnormal period, as well as the association between each abnormal period and each non-abnormal period in the cluster to which it belongs, the degree of failure of each abnormal period is obtained; based on the degree of failure and duration of each abnormal period, the duration of adjacent abnormal periods of each abnormal period, and the time interval between adjacent abnormal periods, the failure probability of the crusher is obtained. Instead of simply monitoring the failure based on the vibration data of the crusher, different types of forestry waste are considered and analyzed from multiple aspects to obtain the accurate failure probability of the crusher, improving the accuracy of crusher failure monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of clustering data processing, and particularly relates to a forestry waste crushing device and method. Background Art

[0002] Forestry waste mainly consists of a type of lignocellulosic waste composed of forest logging residues, wood processing residues, and forest pruning residues. It is a type of potential resource with great potential. By crushing forestry waste, forestry waste can be applied to garden covering, spreading inorganic or organic substances regularly on the soil, which can not only protect the soil but also improve the soil quality. Therefore, crushing is a key step in forestry waste treatment. During the process of forestry waste treatment, a crusher is the main equipment for forestry waste crushing. If a fault occurs in the crusher and is not detected in time, production will be interrupted. Once production is interrupted, subsequent production links will be affected, causing the entire production line to come to a standstill.

[0003] When the prior art monitors the faults of crushers used for forestry waste, it usually directly analyzes the initial vibration data of the crushers and determines whether there are faults in the crushers according to the analysis results. However, in the actual forestry waste treatment environment, forestry waste has differences in many aspects such as hardness, toughness, and size. The vibration patterns generated by different characteristics of forestry waste during the crushing process vary greatly. If the differences between forestry waste are not considered and the fault monitoring is directly carried out based on the vibration data of the crusher, the accuracy of crusher fault monitoring will be affected. Summary of the Invention

[0004] In order to solve the technical problem of the low accuracy of fault monitoring of crushers applied to forestry waste in the prior art, the purpose of the present invention is to provide a forestry waste crushing device and method, and the specific technical solutions adopted are as follows:

[0005] In the first aspect of the present invention, a forestry waste crushing method is provided, including:

[0006] Obtaining multiple vibration data periods and power data periods of the crusher, and clustering the vibration data periods to obtain multiple clusters;

[0007] Selecting abnormal periods by using the probability of fault occurrence of each vibration data period in the cluster, where the probability of fault occurrence is obtained from the clustering effect of the cluster and the correlation between the vibration data period in the cluster and the power data period in the same period;

[0008] Based on the probability of fault occurrence of each abnormal period and the association between each abnormal period and each non-abnormal period in the cluster where it is located, obtaining the degree of fault of each abnormal period;

[0009] Based on the fault degree and duration of each abnormal period, the duration of the adjacent abnormal period of each abnormal period, and the time interval between the adjacent abnormal periods, the fault probability of the crusher is obtained.

[0010] In an exemplary embodiment, the process of obtaining the fault degree includes:

[0011] According to the time interval and eigenvalue difference between the first abnormal period and each non-abnormal period in the cluster where it is located, and combining the probability of failure of the first abnormal period and each non-abnormal period in the cluster where it is located, the initial fault degree of the first abnormal period is obtained; the first abnormal period is any abnormal period.

[0012] According to the initial fault degree of the first abnormal period and the duration ratio of the first abnormal period, the fault degree of the first abnormal period is obtained.

[0013] In an exemplary embodiment, the process of obtaining the initial fault degree of the first abnormal period includes:

[0014] According to the first time interval, the first eigenvalue difference, and the probability of failure of the first non-abnormal period, the fault influence sub-coefficient of the first non-abnormal period on the first abnormal period is obtained; the fault influence sub-coefficient is inversely proportional to the first time interval, directly proportional to the first eigenvalue difference, and inversely proportional to the probability of failure of the first non-abnormal period; the first non-abnormal period is any non-abnormal period in the cluster where the first abnormal period is located; the first time interval is the time interval between the first abnormal period and the first non-abnormal period, and the first eigenvalue difference is the eigenvalue difference between the first abnormal period and the first non-abnormal period.

[0015] Fuse the fault influence sub-coefficients of each non-abnormal period in the cluster where the first abnormal period is located on the first abnormal period to obtain the fault influence coefficient of the first abnormal period.

[0016] According to the fault influence coefficient of the first abnormal period and the probability of failure of the first abnormal period, the initial fault degree of the first abnormal period is obtained.

[0017] In an exemplary embodiment, the process of obtaining the fault probability of the crusher includes:

[0018] According to the duration of the adjacent abnormal period of each abnormal period and the time interval between each abnormal period and the adjacent abnormal period, the fault probability influence weight related to each abnormal period is obtained; the fault probability influence weight is directly proportional to the duration of the adjacent abnormal period of the abnormal period and inversely proportional to the time interval between the abnormal period and the adjacent abnormal period.

[0019] Based on the failure probability impact weights related to each abnormal period, the failure degrees of each abnormal period are weighted and summed, and combined with the mean value of the time intervals between any two abnormal periods, the failure probability of the crusher is obtained.

[0020] In an exemplary embodiment, the process of obtaining the probability of failure includes:

[0021] Obtain the correlation coefficient between the first vibration data period in the first type of cluster and the power data period in the same period; the first type of cluster is any one of the clusters, and the first vibration data period is any one of the vibration data periods in the first type of cluster;

[0022] Take the clustering effect of the first type of cluster as the credibility of failure occurrence of the first type of cluster, and combine the correlation coefficient between the first vibration data period and the power data period in the same period to obtain the probability of failure occurrence of the first vibration data period.

[0023] In an exemplary embodiment, the process of obtaining the clustering effect of the first type of cluster includes: obtaining the clustering effect of the first type of cluster according to the average value of the similarity between any two vibration data periods in the first type of cluster and the silhouette coefficient of the first type of cluster.

[0024] In an exemplary embodiment, before obtaining multiple vibration data periods and power data periods of the crusher, the forestry waste crushing method further includes:

[0025] Obtain the vibration data sequence and power data sequence of the crusher;

[0026] Use the STL algorithm to decompose the vibration data sequence to obtain the trend item curve;

[0027] Use the APCA segmentation method to segment the trend item curve, and based on the segmentation points, segment the vibration data sequence and the power data sequence to obtain multiple vibration data periods and power data periods respectively.

[0028] In an exemplary embodiment, clustering the vibration data periods to obtain multiple clusters includes:

[0029] Obtain the difference between the APCA approximations of any two trend item periods, where the trend item periods are obtained by segmenting the trend item curve using the APCA segmentation method;

[0030] Use the K-means clustering algorithm, take the difference between the APCA approximations of any two trend item periods as the clustering distance, cluster the trend item periods, and based on the relationship between the trend item periods and the vibration data periods, obtain multiple clusters, and the clusters include multiple vibration data periods.

[0031] In an exemplary embodiment, the forestry waste crushing method further includes:

[0032] Compare the failure probability of the crusher with a preset failure threshold;

[0033] If the failure probability of the crusher is greater than the preset failure threshold, output a crusher shutdown instruction.

[0034] In a second aspect of the present invention, there is provided a forestry waste crushing device, comprising: a memory and a processor; the memory is connected to the processor; the memory is used for storing program instructions; the processor is used for implementing the above-mentioned forestry waste crushing method when the program instructions are executed.

[0035] The present invention has the following beneficial effects: clustering the vibration data time periods of the crusher to obtain multiple clusters, and the crushed forestry waste in each cluster is similar, so as to better reflect the similarity between the changes of the crusher vibration data in different time periods and improve the accuracy of crusher failure detection; then, focus on analyzing the selected abnormal time periods. Since there is a certain correlation between each abnormal time period and each non-abnormal time period in the cluster where it is located, and this correlation is related to the failure degree of each abnormal time period, therefore, based on this, the failure degree of each abnormal time period is obtained, realizing taking a single abnormal time period as the analysis object to obtain the accurate failure degree of each abnormal time period. Finally, since there is a certain correlation between an abnormal time period and the adjacent abnormal time period, whether it is the duration or the time interval of the abnormal time period, this correlation affects the failure probability of the crusher. Therefore, the technical solution provided by the present invention does not simply perform failure monitoring based on the vibration data of the crusher, but considers different types of forestry waste, analyzes from multiple aspects, obtains the accurate failure probability of the crusher, and improves the accuracy of crusher failure monitoring. Description of the Drawings

[0036] Figure 1 is a flowchart of a forestry waste crushing method provided by an embodiment of the present invention;

[0037] Figure 2 is a flowchart of the steps further included in the forestry waste crushing method provided by an embodiment of the present invention;

[0038] Figure 3 is a schematic diagram of a curve obtained by decomposing the vibration data fitting curve using the STL algorithm provided by an embodiment of the present invention;

[0039] Figure 4 is a specific clustering process diagram provided by an embodiment of the present invention;

[0040] Figure 5 is a flowchart for obtaining the probability of occurrence of a failure provided by an embodiment of the present invention;

[0041] Figure 6 It is a flowchart for obtaining the degree of fault provided by an embodiment of the present invention;

[0042] Figure 7 It is a flowchart for obtaining the initial fault degree of the first abnormal period provided by an embodiment of the present invention;

[0043] Figure 8 It is a flowchart for obtaining the fault probability of a crusher provided by an embodiment of the present invention;

[0044] Figure 9 It is a flowchart of the steps further included in the forestry waste crushing method provided by an embodiment of the present invention. Detailed implementation manners

[0045] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following, in combination with the accompanying drawings and preferred embodiments, details the specific implementation manners, structures, features and effects of the present invention. In the following description, different "an embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0046] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs. All data information collected in this application is obtained through full consent and authorization, and the collection, use and processing of relevant information need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0047] The application scenario of the forestry waste crushing method provided in this embodiment is as follows: There are many types of forestry waste, with large differences in hardness, toughness, size, etc. During the crushing process, waste of different properties has different vibration trends. Therefore, the abnormality of vibration data in several periods with the same vibration trend is analyzed to improve the accuracy of crusher fault monitoring.

[0048] Such as Figure 1 As shown, the forestry waste crushing method includes the following steps:

[0049] Step S1: Obtain multiple vibration data periods and power data periods of the crusher, and cluster the vibration data periods to obtain multiple clusters;

[0050] Step S2: Screen out abnormal periods based on the probability of failure of each vibration data period in the cluster, and the probability of failure is obtained from the clustering effect of the cluster and the correlation between the vibration data period in the cluster and the power data period in the same period;

[0051] Step S3: Based on the probability of a fault occurring in each abnormal period and the association between each abnormal period and each non-abnormal period in the cluster to which it belongs, obtain the fault degree of each abnormal period;

[0052] Step S4: Based on the fault degree and duration of each abnormal period, the duration of the adjacent abnormal period of each abnormal period, and the time interval between adjacent abnormal periods, obtain the fault probability of the crusher.

[0053] The following will specifically describe each step in conjunction with the accompanying drawings.

[0054] Step S1: Obtain multiple vibration data periods and power data periods of the crusher, and perform clustering on the vibration data periods to obtain multiple clusters.

[0055] Obtain multiple vibration data periods and power data periods of the crusher. The multiple vibration data periods and power data periods of the crusher can be pre-processed and obtained, and are directly used in step S1 of the present invention. In an exemplary embodiment, before obtaining multiple vibration data periods and power data periods of the crusher, as Figure 2 shown, the forestry waste crushing method provided by the present invention further includes:

[0056] Step S1-1: Obtain the vibration data sequence and power data sequence of the crusher.

[0057] In this embodiment, a blade crusher is used as an example for fault detection. The blade crusher usually cuts and crushes forestry waste with blades installed on a rotating shaft. Therefore, the rotating shaft is the core component of the crusher. By transmitting the power of the motor to the crushing tool, the tool can rotate at a high speed to crush forestry waste. In the structure of the forestry waste crusher, the rotating shaft is a position where it is relatively easy to install a vibration sensor. Compared with some other complex components, the surface of the rotating shaft is relatively regular, which is convenient for fixing the sensor and can ensure good contact between the vibration sensor and the rotating shaft, so as to obtain accurate vibration data. Since the vibration of the crusher is caused by the rotation of the rotating shaft, in other embodiments, the vibration sensor can also be fixed on the body of the crusher, and the vibration data of the crusher body detected is used to characterize the vibration condition of the rotating shaft. Moreover, a power sensor is also provided at the power supply end of the crusher to detect the operating power of the crusher.

[0058] The acquisition frequencies of the vibration sensor and the power sensor of the crusher are the same, and the two collect data synchronously. In an exemplary embodiment, the acquisition frequency is once every 5 seconds.

[0059] This embodiment also sets a monitoring time period. In subsequent processing, by analyzing the vibration data and power data of the crusher within the monitoring time period, crusher fault monitoring is performed. The end time of the monitoring time period is the current time, and the duration of the monitoring time period is set according to the actual situation. Then, the monitoring time period includes multiple sampling times, and each sampling time corresponds to a vibration data and a power data. Arrange the vibration data at each collected time in chronological order to obtain the vibration data sequence of the crusher, and arrange the power data at each collected time in chronological order to obtain the power data sequence of the crusher.

[0060] Step S1-2: Decompose the vibration data sequence using the STL algorithm to obtain a trend term curve.

[0061] In the scenario of a forestry waste crusher, the vibration data of the crusher is affected by various factors, including both long-term trend changes brought about by changes in the operating state of the equipment itself and some periodic or random interference factors. By using the STL (Seasonal-Trend decomposition using Loess) algorithm to extract the trend term, it is possible to focus on the long-term change pattern in the vibration data, remove interference such as short-term fluctuations, seasonality, and random noise, and more clearly understand the basic change trend of the rotation shaft vibration.

[0062] Therefore, use the STL algorithm to decompose the vibration data sequence to obtain a trend term curve. Among them, before using the STL algorithm to decompose the vibration data sequence, perform curve fitting on the vibration data sequence to obtain a vibration data fitting curve. The horizontal axis of the vibration data fitting curve is the time axis, and the vertical axis is the amplitude axis of the vibration data. When using the STL algorithm to decompose the vibration data fitting curve, in addition to obtaining the trend term curve, a periodic term and a residual term are also obtained. As Figure 3 shown, the curves from bottom to top are: the residual term curve, the periodic term curve, the trend term curve, and the vibration data fitting curve. Figure 3 The horizontal axis in

[0063] The vibration trend of the rotation shaft of the crusher may have different change rates and characteristics in different stages. For example, when crushing harder branches, the vibration trend of the rotation shaft may be continuous and fluctuate greatly, and the corresponding characteristic values will reflect this higher vibration level; while when crushing softer leaves, the vibration trend is relatively gentle, and the corresponding characteristic values will be lower.

[0064] Step S1-3: Segment the trend term curve using the APCA segmentation method, and based on the segmentation points, segment the vibration data sequence and the power data sequence to obtain multiple vibration data time periods and power data time periods respectively.

[0065] The piecewise method of APCA (Adaptive Piecewise Constant Approximation) is used to segment the trend term curve, obtaining several trend term time periods and the APCA approximation values of each trend term time period. The APCA approximation value can be the data mean in the trend term time period. Correspondingly, multiple segmentation points are obtained. Then, based on the obtained multiple segmentation points, the vibration data sequence is segmented to obtain multiple vibration data time periods, and the power data sequence is segmented to obtain multiple power data time periods. Therefore, the number of vibration data time periods is the same as that of the power data time periods, and each vibration data time period corresponds to each power data time period one by one.

[0066] By segmenting through the APCA piecewise method and obtaining the APCA approximation values, the entire continuous trend change can be divided into multiple representative time periods according to the internal characteristic differences, which is convenient for more detailed capture of the characteristics of vibration trends in different stages.

[0067] Then, clustering is performed on the vibration data time periods to obtain multiple clusters. In an exemplary embodiment, since the APCA approximation values of each of the above-obtained trend term time periods are obtained, then, as Figure 4 shown, a specific clustering process is given as follows:

[0068] Step S1-4: Obtain the difference between the APCA approximation values of any two trend term time periods.

[0069] Obtain the difference between the APCA approximation values of any two trend term time periods, where the difference between the APCA approximation values is specifically the absolute value of the difference between the APCA approximation values.

[0070] Step S1-5: Use the K-means clustering algorithm, with the difference between the APCA approximation values of any two trend term time periods as the clustering distance, perform clustering on the trend term time periods, and based on the relationship between the trend term time periods and the vibration data time periods, obtain multiple clusters, where each cluster includes multiple vibration data time periods.

[0071] In this embodiment, the K-means clustering algorithm is used for clustering, where the value of K is set according to the actual clustering needs. Using the difference between the APCA approximation values of any two trend term time periods as the clustering distance, clustering is performed on the trend term time periods to obtain multiple clusters. Since there is a one-to-one correspondence between each trend term time period and each vibration data time period, therefore, after obtaining the respective clusters corresponding to the trend term time periods, replacing the trend term time periods in each cluster with the vibration data time periods of the same time period, multiple clusters corresponding to the vibration data time periods are obtained, and each cluster includes multiple vibration data time periods, realizing the clustering of the vibration data time periods.

[0072] Therefore, adopting the K-means clustering algorithm to cluster the trend-term periods is essentially to cluster each time period corresponding to the trend-term periods, and the obtained clusters include multiple time periods. Then, according to the corresponding relationship in time between the trend-term periods and the vibration data periods, clusters including multiple vibration data periods are obtained.

[0073] Correspondingly, according to the one-to-one correspondence between the vibration data periods and the power data periods, multiple power data periods included in each cluster are obtained.

[0074] Since the vibration trend is determined by the properties of the forestry waste being crushed, the crushed forestry waste in each cluster has similarities in terms of hardness, toughness, size, etc.

[0075] Step S2: Screen to obtain abnormal periods based on the probability of failure occurrence in each vibration data period of the cluster. The probability of failure occurrence is obtained from the clustering effect of the cluster and the correlation between the vibration data periods and the power data periods in the same time period in the cluster.

[0076] The clustering process in Step S1 only considers the change of the trend term. In this step, according to the fluctuation difference between the vibration data, it is determined whether the clustering division is appropriate. Whether the clustering division is appropriate is characterized by the clustering effect of each cluster.

[0077] Set the first cluster as any cluster. Based on the clustering process in Step S1, the silhouette coefficient of the first cluster is obtained. The larger the silhouette coefficient, the better the clustering effect of the first cluster. Moreover, since the first cluster includes multiple vibration data periods, therefore, obtain the similarity between any two vibration data periods in the first cluster. The specific implementation manner of the similarity is determined according to actual needs. In an exemplary embodiment, obtain the DTW (Dynamic Time Warping Distance) distance between any two vibration data periods in the first cluster, and then perform negative correlation normalization on the DTW distance to obtain the similarity between any two vibration data periods in the first cluster. The negative correlation normalization method here can be: , where exp is the exponential function with the natural constant e as the base, and x is the object to be negatively correlated and normalized.

[0078] Then calculate the average value of the similarity between any two vibration data periods in the first cluster. Then, the higher the average value of the similarity, the better the clustering effect of the first cluster.

[0079] Therefore, according to the average value of the similarity between any two vibration data periods in the first type of cluster and the silhouette coefficient of the first type of cluster, the clustering effect of the first type of cluster is obtained. The clustering effect is proportional to the silhouette coefficient and proportional to the average value of the similarity. In an exemplary embodiment, for the convenience of subsequent calculations, the silhouette coefficient of the first type of cluster is normalized. Since the numerical range of the silhouette coefficient of the first type of cluster is [-1, 1], therefore, the silhouette coefficient of the first type of cluster is added with the value 1 and then divided by 2, so that the numerical range of the obtained result is [0, 1], achieving normalization.

[0080] Multiply the average value of the similarity between any two vibration data periods in the first type of cluster by the normalized silhouette coefficient of the first type of cluster as the clustering effect of the first type of cluster.

[0081] The hardness characteristics of forestry waste will simultaneously affect the vibration of the rotating shaft and the power of the crusher. Hard materials will increase power consumption and vibration amplitude. Moreover, the power is also affected by voltage fluctuations and motor failures, that is, voltage fluctuations in the power supply system will affect the operating efficiency of the motor, thereby affecting the power stability of the crusher. Motor failures or aging may also cause unstable power output. Therefore, the smaller the correlation between the vibration of the rotating shaft and the power of the crusher, the more likely a failure will occur. Therefore, according to the clustering effect of the first type of cluster and the correlation between the vibration data period and the power data period in the same period in the first type of cluster, the probability of failure of the first type of cluster is obtained.

[0082] In an exemplary embodiment, as Figure 5 shown, the process of obtaining the probability of failure includes:

[0083] Step S2-1: Obtain the correlation coefficient between the first vibration data period and the power data period in the same period in the first type of cluster.

[0084] For each vibration data period in the first type of cluster, there is a corresponding power data period in the same period. For the convenience of description, the first vibration data period is set as any one vibration data period in the first type of cluster.

[0085] Obtain the correlation coefficient between the first vibration data period and the power data period in the same period. In an exemplary embodiment, the correlation coefficient between the first vibration data period and the power data period in the same period is specifically the Pearson correlation coefficient between the first vibration data period and the power data period in the same period. The smaller the Pearson correlation coefficient, the more likely an anomaly will occur.

[0086] Step S2-2: Use the clustering effect of the first type of cluster as the confidence level of failure occurrence of the first type of cluster, and combine it with the correlation coefficient between the first vibration data period and the power data period in the same period to obtain the probability of failure of the first vibration data period.

[0087] Take the clustering effect of the first type of cluster as the fault occurrence credibility of the first type of cluster. The greater the fault occurrence credibility, the more similar the forestry waste in different time periods within the first type of cluster, and the more credible the possibility of a fault occurring.

[0088] Based on the fault occurrence credibility of the first type of cluster and the correlation coefficient between the first vibration data period and the power data period in the same time period, obtain the possibility of a fault occurring in the first vibration data period. In an exemplary embodiment, normalize the negative correlation between the first vibration data period and the power data period in the same time period for subsequent calculations. Since the value range of the Pearson correlation coefficient is [-1, 1], subtract the Pearson correlation coefficient from the value 1 and then divide by 2, so that the value range of the obtained result is [0, 1], achieving negative correlation normalization.

[0089] Then multiply the normalized correlation coefficient between the first vibration data period and the power data period in the same time period by the clustering effect of the first type of cluster, and take the obtained result as the possibility of a fault occurring in the first vibration data period. In this way, obtain the possibility of a fault occurring in each vibration data period of each cluster.

[0090] Use the possibility of a fault occurring in each vibration data period in the first type of cluster to screen out the abnormal periods in the first type of cluster. The abnormal periods are specifically abnormal vibration data periods. In an exemplary embodiment, preset an abnormal threshold, and the value range of this preset abnormal threshold is 0 - 1, and the specific value of this preset abnormal threshold is set according to the actual situation. If a safer monitoring logic is required, this preset abnormal threshold can be set slightly smaller, so that it is easier to have abnormal periods that meet the monitoring requirements. In this embodiment, the preset abnormal threshold is taken as an example of 0.6. Then, compare the possibility of a fault occurring in each vibration data period in the first type of cluster with this preset abnormal threshold, and determine the vibration data period corresponding to the possibility of a fault occurring that is greater than this preset abnormal threshold as the abnormal period. Then, through screening, the first type of cluster includes abnormal periods and non-abnormal periods (i.e., normal periods).

[0091] Adopt the above process to obtain the abnormal periods in each cluster. Then, overall, obtain multiple abnormal periods.

[0092] Step S3: Based on the possibility of a fault occurring in each abnormal period and the association between each abnormal period and each non-abnormal period in the cluster where it is located, obtain the fault degree of each abnormal period.

[0093] The abnormal period represents the period when the crusher may malfunction. Due to the complex working environment of the forestry waste crusher and the wide variety of forestry wastes, there are significant differences in hardness, water content, etc. Under long-term high-load operation, different types of faults may occur, but the impact degrees of different types of faults are different. It is necessary to further determine the type of fault and the fault degree of the crusher according to the distribution characteristics of the abnormal periods in time series.

[0094] In an exemplary embodiment, as Figure 6 shown, the process of obtaining the fault degree includes:

[0095] Step S3-1: According to the time interval and eigenvalue difference between the first abnormal period and each non-abnormal period in the cluster it belongs to, and combining the fault occurrence probabilities of the first abnormal period and each non-abnormal period in the cluster it belongs to, obtain the initial fault degree of the first abnormal period.

[0096] For each obtained abnormal period, sort the abnormal periods according to the chronological order of each abnormal period, so as to obtain the chronological order of each abnormal period in time series.

[0097] For the sake of convenience of explanation, set the first abnormal period as any abnormal period, then the cluster to which the first abnormal period belongs and each non-abnormal period in the cluster to which the first abnormal period belongs can be obtained.

[0098] According to the time interval and eigenvalue difference between the first abnormal period and each non-abnormal period in the cluster it belongs to, and combining the fault occurrence probabilities of the first abnormal period and each non-abnormal period in the cluster it belongs to, obtain the initial fault degree of the first abnormal period.

[0099] In an exemplary embodiment, as Figure 7 shown, the following gives the process of obtaining the initial fault degree of the first abnormal period, including:

[0100] Step S3-1-1: According to the first time interval, the first eigenvalue difference, and the fault occurrence probability of the first non-abnormal period, obtain the fault influence sub-coefficient of the first non-abnormal period on the first abnormal period.

[0101] For the sake of convenience of explanation, set the first non-abnormal period as any non-abnormal period in the cluster to which the first abnormal period belongs.

[0102] Obtain the time interval between the first abnormal period and the first non-abnormal period. Specifically: If the first abnormal period is earlier than the first non-abnormal period, then obtain the end time of the first abnormal period and the start time of the first non-abnormal period, and take the time interval between the end time of the first abnormal period and the start time of the first non-abnormal period as the time interval between the first abnormal period and the first non-abnormal period; If the first non-abnormal period is earlier than the first abnormal period, then obtain the end time of the first non-abnormal period and the start time of the first abnormal period, and take the time interval between the end time of the first non-abnormal period and the start time of the first abnormal period as the time interval between the first abnormal period and the first non-abnormal period. For the convenience of description, the first time interval is defined as the time interval between the first abnormal period and the first non-abnormal period.

[0103] Obtain the characteristic value of the first abnormal period and the characteristic value of the first non-abnormal period. The characteristic value represents the characteristic situation of the corresponding period. In an exemplary embodiment, based on the above APCA segmentation method, the characteristic value is specifically the APCA approximation value of the corresponding period, so as to obtain the APCA approximation value of the first abnormal period and the APCA approximation value of the first non-abnormal period. And obtain the characteristic value difference between the first abnormal period and the first non-abnormal period. The characteristic value difference is specifically the absolute value of the difference between the APCA approximation value of the first abnormal period and the APCA approximation value of the first non-abnormal period, and it is defined as the first characteristic value difference.

[0104] According to the first time interval, the first characteristic value difference, and the probability of failure occurrence in the first non-abnormal period, obtain the failure influence sub-coefficient of the first non-abnormal period on the first abnormal period. The longer the first time interval, the lower the credibility of the failure influence sub-coefficient; the shorter the first time interval, the higher the credibility of the failure influence sub-coefficient. The greater the first characteristic value difference, the greater the difference in vibration performance between the first abnormal period and the first non-abnormal period, and the greater the probability of failure occurrence. The smaller the probability of failure occurrence in the first non-abnormal period, the more credible the first characteristic value difference. Therefore, the failure influence sub-coefficient is inversely proportional to the first time interval, directly proportional to the first characteristic value difference, and inversely proportional to the probability of failure occurrence in the first non-abnormal period.

[0105] In an exemplary embodiment, the calculation formula of the failure influence sub-coefficient is as follows:

[0106] ;

[0107] Among them, represents the failure influence sub-coefficient of the z-th non-abnormal period in the cluster where the y-th abnormal period is located on the y-th abnormal period, represents the probability of failure occurrence of the z-th non-abnormal period in the cluster where the y-th abnormal period is located, represents the time interval between the y-th abnormal period and the z-th non-abnormal period in the cluster where it is located, represents the APCA approximation value of the y-th abnormal period, represents the APCA approximation value of the z-th non-abnormal period in the cluster where the y-th abnormal period is located.

[0108] represents the normalization of , and the normalization method here can be: .

[0109] represents the normalization of the product of three parts. The normalization method here can be: obtain the maximum and minimum values of the product of the three parts of the y-th abnormal period by each non-abnormal period in the cluster where the y-th abnormal period is located, so as to normalize the product of the three parts based on the maximum-minimum value normalization method.

[0110] The larger , the greater the difference in vibration performance between the y-th abnormal period and the z-th non-abnormal period in the cluster where the y-th abnormal period is located, and the greater the possibility of failure; the time interval The smaller , the higher the credibility of the fault influence sub-coefficient; the possibility of failure The smaller , the corresponding is more credible.

[0111] Step S3-1-2: Fuse the fault influence sub-coefficients of each non-abnormal period in the cluster where the first abnormal period is located on the first abnormal period to obtain the fault influence coefficient of the first abnormal period.

[0112] Step S3-1-1 obtains the fault influence sub-coefficients of each non-abnormal period in the cluster where the y-th abnormal period is located on the y-th abnormal period. Fuse the fault influence sub-coefficients of each non-abnormal period in the cluster where the y-th abnormal period is located on the y-th abnormal period to obtain the fault influence coefficient of the y-th abnormal period. In an exemplary embodiment, calculate the average value of the fault influence sub-coefficients of each non-abnormal period in the cluster where the y-th abnormal period is located on the y-th abnormal period as the fault influence coefficient of the y-th abnormal period.

[0113] Step S3-1-3: Obtain the initial fault degree of the first abnormal period according to the fault influence coefficient of the first abnormal period and the possibility of failure of the first abnormal period.

[0114] Calculate the product of the fault influence coefficient of the y-th abnormal period and the possibility of failure of the y-th abnormal period as the initial fault degree of the y-th abnormal period.

[0115] Step S3-2: Obtain the fault degree of the first abnormal period according to the initial fault degree of the first abnormal period and the duration ratio of the first abnormal period.

[0116] Obtain the duration of the y-th abnormal period, and obtain the total duration of all abnormal periods. Calculate the ratio of the duration of the y-th abnormal period to the total duration as the duration ratio of the y-th abnormal period. The larger the duration ratio of the y-th abnormal period, the more serious the fault that occurs in the y-th abnormal period, and the greater the fault degree.

[0117] Calculate the product of the initial fault degree of the y-th abnormal period and the duration ratio of the y-th abnormal period as the fault degree of the y-th abnormal period.

[0118] Adopt the above process to obtain the fault degree of each abnormal period.

[0119] Step S4: Based on the fault degree and duration of each abnormal period, the duration of the adjacent abnormal period of each abnormal period, and the time interval between adjacent abnormal periods, obtain the fault probability of the crusher.

[0120] The more dispersed the abnormal periods are in time series, it may be caused by unstable power supply voltage. For example, a short-term voltage increase or decrease will cause changes in parameters such as the speed and torque of the motor, thus causing abnormal vibration of the crusher. And after the abnormality occurs, it will return to normal immediately, and the crusher is affected relatively briefly, and the degree of fault is relatively light. But when the abnormal periods are relatively continuous in time series, it means that the crusher is affected by continuous interference, and the crusher may have more serious faults, such as damage to crushing tools (such as hammer pieces, blades, etc.), bearing failures, etc., which will not only affect the waste crushing effect, but also may gradually increase the degree of fault as the crusher operates. Therefore, the more continuous the abnormal periods are in time series, the greater the probability of the crusher failing and the higher the degree of fault.

[0121] Therefore, based on the fault degree and duration of each abnormal period, the duration of the adjacent abnormal period of each abnormal period, and the time interval between adjacent abnormal periods, obtain the fault probability of the crusher. In an exemplary embodiment, as Figure 8 shown, the process of obtaining the fault probability of the crusher includes:

[0122] Step S4-1: According to the duration of the adjacent abnormal period of each abnormal period and the time interval between each abnormal period and the adjacent abnormal period, obtain the fault probability influence weight related to each abnormal period.

[0123] Obtain the adjacent abnormal periods of the y-th abnormal period in time sequence. Among them, if the y-th abnormal period is the first abnormal period in time sequence, then the abnormal period on the right (i.e., the second abnormal period in time sequence) is used as its adjacent abnormal period; if the y-th abnormal period is the last abnormal period in time sequence, then the abnormal period on the left (i.e., the second-to-last abnormal period in time sequence) is used as its adjacent abnormal period; if the y-th abnormal period is any abnormal period from the second abnormal period to the second-to-last abnormal period in time sequence, then one abnormal period on each of its left and right sides is used as its adjacent abnormal periods.

[0124] Obtain the duration of the y-th abnormal period and the durations of its adjacent abnormal periods. It should be understood that if there are adjacent abnormal periods on both the left and right sides of the y-th abnormal period, then the average of the duration of the left adjacent abnormal period and the duration of the right adjacent abnormal period of the y-th abnormal period is obtained as the duration of the adjacent abnormal period of the y-th abnormal period.

[0125] Obtain the time interval between the y-th abnormal period and its adjacent abnormal periods. It should be understood that if there are adjacent abnormal periods on both the left and right sides of the y-th abnormal period, then the average of the time interval between the y-th abnormal period and the left adjacent abnormal period and the time interval between the y-th abnormal period and the right adjacent abnormal period is obtained as the time interval between the y-th abnormal period and its adjacent abnormal periods.

[0126] According to the duration of the adjacent abnormal period of the y-th abnormal period and the time interval between the y-th abnormal period and its adjacent abnormal periods, obtain the fault probability influence weight related to the y-th abnormal period. Among them, the shorter the time interval between the y-th abnormal period and its adjacent abnormal periods, the more continuous the y-th abnormal period is in time sequence, and the higher the fault probability; the longer the duration of the adjacent abnormal period of the y-th abnormal period, the more continuous the y-th abnormal period is in time sequence, and the higher the fault probability. Therefore, the fault probability influence weight corresponding to the y-th abnormal period is directly proportional to the duration of the adjacent abnormal period of the y-th abnormal period and inversely proportional to the time interval between the y-th abnormal period and its adjacent abnormal periods.

[0127] In an exemplary embodiment, the calculation formula for the fault probability influence weight corresponding to the y-th abnormal period is as follows:

[0128] ;

[0129] Among them, represents the fault probability influence weight corresponding to the y-th abnormal period, represents the duration of the adjacent abnormal period of the y-th abnormal period, To show the time interval between the y-th abnormal period and the adjacent abnormal periods, where Y represents the number of abnormal periods. Adding 0.01 to the denominator is to avoid the denominator being zero, thus ensuring the denominator is meaningful.

[0130] Through The calculation formula, the sum of the fault probability influence weights corresponding to all abnormal periods is 1.

[0131] Step S4-2: Based on the fault probability influence weights related to each abnormal period, perform a weighted sum of the fault degrees of each abnormal period, and combine the mean value of the time intervals between any two abnormal periods to obtain the fault probability of the crusher.

[0132] According to the obtained fault probability influence weights corresponding to each abnormal period and the fault degrees of each abnormal period, use the fault probability influence weights corresponding to each abnormal period as the weight coefficients of the fault degrees of the corresponding abnormal periods, and perform a weighted sum of the fault degrees of each abnormal period. The calculation formula is:

[0133] ;

[0134] Among them, Represents the result after performing a weighted sum of the fault degrees of each abnormal period, which is the weighted fault degree of the crusher. The higher the fault degree, the higher the fault probability of the crusher. Represents the fault degree of the y-th abnormal period.

[0135] Obtain the time interval between any two abnormal periods, and calculate the average value of the time intervals between any two abnormal periods as the overall interval situation of the abnormal periods. The smaller the average value of the time intervals, the more continuous the abnormal periods are. Correspondingly, the higher the fault probability of the crusher.

[0136] According to the result of the weighted sum of the fault degrees of each abnormal period and the average value of all the time intervals between any two abnormal periods, obtain the fault probability of the crusher. In an exemplary embodiment, the calculation formula of the fault probability of the crusher is given as follows:

[0137] ;

[0138] Among them, Represents the fault probability of the crusher, Represents the average value of all the time intervals between any two abnormal periods.

[0139] Thus, the fault probability of the crusher is obtained. The higher the fault probability, the more likely the crusher has a fault.

[0140] In an exemplary embodiment, such as Figure 9As shown, the forestry waste crushing method further includes:

[0141] Step S5: Compare the failure probability of the crusher with a preset failure threshold;

[0142] Step S6: If the failure probability of the crusher is greater than the preset failure threshold, output a crusher shutdown instruction.

[0143] Among them, the value range of the preset failure threshold is 0 - 1, and the specific value of this preset failure threshold is set according to the actual situation. If a safer monitoring logic is required, this preset failure threshold can be set slightly smaller, so that it is easier to obtain judgment results that meet the monitoring requirements. In this embodiment, the preset abnormal threshold is taken as 0.6 for example. The crusher stops running under the action of the crusher shutdown instruction, avoiding more serious accidents.

[0144] This embodiment also provides a forestry waste crushing device, including: a memory and a processor; the memory is connected to the processor, and the memory is used to store program instructions; the processor is used to implement the steps in the above-mentioned forestry waste crushing method embodiment when the program instructions are executed.

[0145] In an exemplary embodiment, the present invention provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, it implements the steps in the above-mentioned forestry waste crushing method embodiment.

[0146] It should be noted that: the above-mentioned sequence of the embodiments of the present invention is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0147] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A method for crushing forestry waste, characterized in that, Including: Obtaining multiple vibration data periods and power data periods of a crusher, and clustering the vibration data periods to obtain multiple clusters; Filtering to obtain abnormal periods based on the probability of failure occurrence of each vibration data period in the cluster, where the probability of failure occurrence is obtained from the clustering effect of the cluster and the correlation between the vibration data period and the power data period in the same period in the cluster; Among them, the process of obtaining the probability of failure occurrence is as follows: obtaining the correlation coefficient between the first vibration data period in the first cluster and the power data period in the same period; the first cluster is any cluster, and the first vibration data period is any vibration data period in the first cluster; taking the clustering effect of the first cluster as the confidence level of failure occurrence of the first cluster, and combining the correlation coefficient between the first vibration data period and the power data period in the same period, to obtain the probability of failure occurrence of the first vibration data period; Based on the probability of failure occurrence of each abnormal period and the association between each abnormal period and each non-abnormal period in the cluster where it is located, obtaining the degree of failure of each abnormal period; Among them, the process of obtaining the degree of failure is as follows: according to the time interval and eigenvalue difference between the first abnormal period and each non-abnormal period in the cluster where it is located, and combining the probability of failure occurrence of the first abnormal period and each non-abnormal period in the cluster where it is located, obtaining the initial degree of failure of the first abnormal period; the first abnormal period is any abnormal period; according to the initial degree of failure of the first abnormal period and the duration ratio of the first abnormal period, obtaining the degree of failure of the first abnormal period; Based on the degree of failure and duration of each abnormal period, the duration of the adjacent abnormal period of each abnormal period, and the time interval between adjacent abnormal periods, obtaining the failure probability of the crusher; Among them, the process of obtaining the failure probability is as follows: according to the duration of the adjacent abnormal period of each abnormal period and the time interval between each abnormal period and the adjacent abnormal period, obtaining the failure probability influence weight related to each abnormal period; the failure probability influence weight is directly proportional to the duration of the adjacent abnormal period of the abnormal period and inversely proportional to the time interval between the abnormal period and the adjacent abnormal period; based on the failure probability influence weights related to each abnormal period, performing weighted summation on the degree of failure of each abnormal period, and combining the mean value of the time intervals between any two abnormal periods, obtaining the failure probability of the crusher.

2. The method for crushing forestry waste according to claim 1, wherein, The process of obtaining the initial degree of failure of the first abnormal period includes: According to the first time interval, the first eigenvalue difference, and the probability of failure occurrence of the first non-abnormal period, obtaining the failure influence sub-coefficient of the first non-abnormal period on the first abnormal period; the failure influence sub-coefficient is inversely proportional to the first time interval, directly proportional to the first eigenvalue difference, and inversely proportional to the probability of failure occurrence of the first non-abnormal period; the first non-abnormal period is any non-abnormal period in the cluster where the first abnormal period is located; the first time interval is the time interval between the first abnormal period and the first non-abnormal period, and the first eigenvalue difference is the eigenvalue difference between the first abnormal period and the first non-abnormal period; Fuse the fault influence sub - coefficients of each non - abnormal period in the cluster where the first abnormal period is located to obtain the fault influence coefficient of the first abnormal period; Based on the fault influence coefficient of the first abnormal period and the probability of the first abnormal period occurring a fault, obtain the initial fault degree of the first abnormal period.

3. The method for crushing forestry waste according to claim 1, characterized in that, The process of obtaining the clustering effect of the first cluster includes: obtaining the clustering effect of the first cluster according to the average value of the similarity between any two vibration data periods in the first cluster and the silhouette coefficient of the first cluster.

4. The method for crushing forestry waste as described in claim 1, characterized in that, Before obtaining multiple vibration data periods and power data periods of the crusher, the forestry waste crushing method further includes: Obtain the vibration data sequence and power data sequence of the crusher; Use the STL algorithm to decompose the vibration data sequence to obtain the trend item curve; Use the APCA segmentation method to segment the trend item curve, and based on the segmentation points, segment the vibration data sequence and power data sequence to obtain multiple vibration data periods and power data periods respectively.

5. The method for crushing forestry waste according to claim 4, characterized in that, Cluster the vibration data periods to obtain multiple clusters, including: Obtain the difference between the APCA approximations of any two trend item periods, where the trend item periods are obtained by segmenting the trend item curve using the APCA segmentation method; Use the K - means clustering algorithm, with the difference between the APCA approximations of any two trend item periods as the clustering distance, cluster the trend item periods, and based on the relationship between the trend item periods and the vibration data periods, obtain multiple clusters, and each cluster includes multiple vibration data periods.

6. The method for crushing forestry waste according to claim 1, characterized in that, The forestry waste crushing method further includes: Compare the fault probability of the crusher with a preset fault threshold; If the fault probability of the crusher is greater than the preset fault threshold, output a crusher shutdown instruction.

7. A forestry waste crushing device, characterized by comprising: A memory and a processor; The memory is connected to the processor; The memory is used to store program instructions; The processor is used to implement the forestry waste crushing method according to any one of claims 1 - 6 when the program instructions are executed.

Images