A self-diagnosis and early warning system for the running state of a scraper

Abstract

The application relates to a scraper operation state self-diagnosis and early warning system and belongs to the technical field of scraper operation monitoring. The system comprises the following: a state acquisition module used for acquiring multiple types of state data of a scraper within a preset time period before the current time; a feature analysis module used for performing time-frequency domain decomposition processing on each type of state data to obtain state features of each type of state data; a weight assignment module used for matching each state feature with a preset abnormal weight table to obtain a matching abnormal parameter type corresponding to each type of state feature and an initial weight of the matching abnormal parameter type; an abnormality judgment module used for adjusting the initial weight of each matching abnormal parameter type to obtain a final weight, calculating a state value of each potential operation abnormality, and screening all operation abnormality states of the scraper at the current time; and an early warning module used for early warning each operation abnormality state. The application can realize accurate diagnosis and early warning of the operation state of the scraper.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of scraper operation monitoring, and in particular to a scraper operation state self-diagnosis and early warning system. BACKGROUND

[0002] As a key equipment for material conveying in mining, metallurgy and other industries, the stability of the scraper operation state directly affects the production efficiency and safety. In the traditional scraper operation maintenance, the state of the scraper is mainly judged by manual regular inspection or single parameter monitoring.

[0003] However, manual inspection has obvious shortcomings. Since the scraper structure is complex, potential faults in key parts such as transmission components, chute interfaces and chain wheel areas are difficult to be intuitively observed by manual inspection, and the inspection cycle is fixed, which cannot capture sudden abnormalities in the scraper operation in real time, resulting in late fault discovery, often causing significant damage, and increasing maintenance costs and downtime.

[0004] The single parameter monitoring method usually detects only a single indicator such as vibration or temperature, and cannot comprehensively reflect the comprehensive operation state of the scraper. When the scraper fails, the change of the single parameter may not be significant, which may cause missed or false judgments. For example, early wear of the transmission component may be accompanied by vibration abnormalities and temperature rise, and monitoring of vibration alone may not timely detect potential changes in temperature, and vice versa, which cannot realize real-time, comprehensive and accurate diagnosis and early warning of the scraper operation state, and cannot meet the needs of modern industry for intelligent maintenance of the scraper. SUMMARY

[0005] To solve the above technical problems, the present application provides a scraper operation state self-diagnosis and early warning system. The technical scheme of the present application is as follows:

[0006] A scraper operation state self-diagnosis and early warning system, comprising:

[0007] A state acquisition module for acquiring a plurality of types of state data of the scraper within a preset time period before the current time through a plurality of types of sensors arranged in advance on the scraper;

[0008] A feature analysis module for performing time-frequency domain decomposition processing on each type of state data to obtain state features of each type of state data;

[0009] A weight assignment module for matching each type of state feature with a preset abnormal weight table to obtain a matching abnormal parameter type and its initial weight corresponding to each type of state feature; the preset abnormal weight table stores a corresponding relationship between abnormal parameter types and initial weights;

[0010] Anomaly judgment module, configured to adjust an initial weight of each matching abnormal parameter type according to a state feature of the matching abnormal parameter type, obtain a final weight, and calculate a state value of each potential running anomaly according to each matching abnormal parameter type, the final weight and the state feature of the matching abnormal parameter type, and filter all running anomaly states of the scraper at the current time from all potential running anomalies according to the state value of each potential running anomaly.

[0011] Early warning module, configured to perform early warning on each running anomaly state according to the state value of each running anomaly state of the scraper at the current time.

[0012] Preferably, the feature analysis module comprises:

[0013] An energy distribution acquisition unit, configured to decompose state data of each type in a preset time period into a plurality of unit time windows through a sliding time window, and acquire time-frequency energy distribution of each unit time window according to Fourier transform;

[0014] A feature extraction unit, configured to acquire frequency features and time domain features of state data of each type at each time node according to the time-frequency energy distribution of each unit time window of each type, wherein the frequency features and the time domain features of each type constitute state features of state data of each type.

[0015] Preferably, the weight assignment module comprises:

[0016] A matching unit, configured to calculate a matching degree between a parameter type name of each type of state features and an abnormal type name of each abnormal parameter type in a preset abnormal weight table through a text semantic matching technology;

[0017] An assignment unit, configured to judge whether the matching degree between each parameter type name and each abnormal type name is greater than a matching degree threshold value, and if the matching degree between any parameter type name and any abnormal type name is greater than the matching degree threshold value, it is determined that the parameter type name and the abnormal type name are matched successfully, and the abnormal parameter type corresponding to the abnormal type name and the initial weight thereof are taken as a matching abnormal parameter type and an initial weight of the state features corresponding to the parameter type name.

[0018] Preferably, the matching unit comprises:

[0019] A vector splitting subunit, configured to convert the parameter type name and the abnormal type name into a state sentence vector and an abnormal sentence vector respectively according to a language vector model;

[0020] A matching degree calculation subunit, configured to calculate a matching degree between each state sentence vector and each abnormal sentence vector.

[0021] Preferably, the matching degree calculation sub-unit realizes formula (1) when calculating the matching degree MD(a, b) between the state sentence vector a and the abnormal sentence vector b:

[0022] (1);

[0023] In formula (1), represents the Euclidean distance, log( ) represents the logarithmic function, N( ) represents the number total statistical function, N(a) represents the number of elements of the state sentence vector a, and N(b) represents the number of elements of the abnormal sentence vector b.

[0024] Preferably, the abnormality judgment module comprises:

[0025] a weight adjustment unit, configured to calculate a distribution index of each matching abnormal parameter type according to a state feature of the matching abnormal parameter type, and adjust an initial weight of the matching abnormal parameter type according to the distribution index to obtain a final weight of the matching abnormal parameter type;

[0026] a state value judgment unit, configured to calculate a state value of each potential running abnormality according to the final weight, the state feature of each matching abnormal parameter type, and a preset risk level of each potential running abnormality, and filter, from all potential running abnormalities, a potential running abnormality with a state value higher than an abnormal state threshold value as all running abnormality states of the scraper at the current moment.

[0027] Preferably, the weight adjustment unit comprises:

[0028] an abnormal matrix construction sub-unit, configured to construct an abnormal matrix according to the state feature of each matching abnormal parameter type;

[0029] a distribution index calculation sub-unit, configured to input the abnormal matrix into a vector space, calculate a vector distance of the state feature of each matching abnormal parameter type in the abnormal matrix based on the vector space, and perform probability conversion on the vector distance to obtain the distribution index of each matching abnormal parameter type;

[0030] a final weight calculation sub-unit, configured to superimpose the distribution index of each matching abnormal parameter type on an initial weight of the matching abnormal parameter type to obtain an activation weight of the matching abnormal parameter type, and perform normalization processing on the activation weights of all matching abnormal parameter types to obtain the final weight of each matching abnormal parameter type.

[0031] Preferably, the distribution index calculation sub-unit realizes formula (2) when calculating the vector distance D of the state feature of any matching abnormal parameter type in the abnormal matrix based on the vector space:

[0032] (2);

[0033] In formula (2), R represents a feature vector corresponding to a state feature of the matched abnormal parameter type, X represents an abnormal matrix, represents a Euclidean distance, T represents a matrix transpose, represents a sum state feature vector obtained by summing feature vectors corresponding to state features of each type in the abnormal matrix, and E represents a standard matrix.

[0034] Preferably, the early warning module comprises:

[0035] An abnormality level judging unit is configured to judge an abnormality level of each running abnormal state according to a state value of each running abnormal state of the scraper at the current time.

[0036] An early warning unit is configured to perform early warning on each running abnormal state according to the abnormality level of each running abnormal state.

[0037] All the optional technical solutions described above can be combined arbitrarily, and the present application does not provide a detailed description of the structure after combination.

[0038] Through the above scheme, the present application has the following beneficial effects:

[0039] The state acquisition module acquires state data of multiple types of the scraper, and the feature analysis module performs time-frequency domain decomposition processing on the state data of each type, so that subsequent analysis of all running abnormal states of the scraper is based on multi-dimensional indexes, which can avoid missed judgment or misjudgment.

[0040] The weight assignment module assigns a matching abnormal parameter type and an initial weight corresponding to a state feature of each type, and the abnormality judging module adjusts the initial weight to obtain a final weight, and then determines all running abnormal states of the scraper according to each matching abnormal parameter type and the state feature and the final weight, and the early warning module performs early warning on each running abnormal state, thereby providing an abnormal state analysis mechanism that can flexibly adjust according to real-time state, which can effectively improve the accuracy of the running abnormal state judgment of the scraper, realize real-time, comprehensive and accurate diagnosis and early warning of the scraper, and meet the demand of intelligent maintenance of the scraper in modern industry.

[0041] The above description is only a summary of the technical solutions of the present application, in order to more clearly understand the technical means of the present application, and the content of the specification can be implemented. The following describes the preferred embodiments of the present application in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 is a structural schematic diagram of a scraper running state self-diagnosis and early warning system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0043] The following examples are intended to illustrate the present application but not to limit the scope of the present application.

[0044] As shown in the drawings, Figure 1 The embodiment of the present application provides a scraper running state self-diagnosis and early warning system, which comprises:

[0045] A state acquisition module is configured to acquire multiple types of state data of the scraper within a preset time period before the current time through multiple types of sensors arranged in advance on the scraper.

[0046] A feature analysis module is configured to perform time-frequency domain decomposition processing on each type of state data to obtain state features of each type of state data.

[0047] A weight assignment module is configured to match each type of state feature with a preset abnormal weight table to obtain a matching abnormal parameter type corresponding to each type of state feature and an initial weight of the matching abnormal parameter type; and the preset abnormal weight table stores a corresponding relationship between abnormal parameter types and initial weights.

[0048] An abnormality judgment module is configured to adjust the initial weight of each matching abnormal parameter type according to the state feature of each matching abnormal parameter type to obtain a final weight, calculate a state value of each potential running abnormality according to each matching abnormal parameter type, the final weight and the state feature, and screen all running abnormality states of the scraper at the current time from all potential running abnormalities according to the state value of each potential running abnormality.

[0049] An early warning module is configured to perform early warning on each running abnormality state according to the state value of each running abnormality state of the scraper at the current time.

[0050] Specifically, in the state acquisition module, the multiple types of sensors can include temperature sensors, vibration sensors, current sensors, voltage sensors, position sensors and strain sensors, etc. Correspondingly, the multiple types of state data include temperature, vibration, current, voltage, position and strain, etc. The specific value of the preset time period can be set as needed, and the preset time period is preferably set to 30 min in the embodiment. It should be noted that after the state data is acquired through the multiple types of sensors, the multiple types of state data are also subjected to time alignment processing for ease of analysis.

[0051] In the feature analysis module, the time-frequency domain decomposition processing is to analyze the state data in both time and frequency dimensions, specifically, to extract features in both time and frequency dimensions for each type of state data, and then to combine the time domain features and the frequency features to obtain the state features of each type.

[0052] In the weight assignment module, the preset abnormal weight table is a table for storing abnormal parameter types and their corresponding initial weights, as shown in Table 1.

[0053] Table 1: Preset abnormal weight table

[0054] Abnormal parameter type Initial weight Tension anomaly 0.25 High temperature anomaly 0.20 Current anomaly 0.15 Vibration deformation 0.12 Position anomaly 0.10 Voltage overload anomaly 0.09 Speed anomaly fault 0.09

[0055] In the abnormality judgment module, potential running abnormalities of the scraper include motor failure, belt failure, scraper failure, track damage, vibration abnormality, etc. The state value is a numerical value for quantifying the severity of the potential running abnormality. The running abnormality state is an abnormal running state whose state value meets the condition from all potential running abnormalities.

[0056] In a specific embodiment, the feature analysis module comprises:

[0057] The energy distribution acquisition unit is configured to decompose the state data of each type in the preset time period into a plurality of unit time windows through a sliding time window, and acquire the time-frequency energy distribution of each unit time window according to Fourier transform.

[0058] The feature extraction unit is configured to acquire the frequency feature and the time domain feature of each time node of the state data of each type according to the time-frequency energy distribution of each unit time window of each type, wherein the frequency feature and the time domain feature of each type constitute the state feature of the state data of each type.

[0059] Specifically, in the energy distribution acquisition unit, the step length of the sliding time window can be set as needed, for example, the step length is set to 3s. After determining the step length of the sliding time window, the state data of each type in the preset time period is decomposed into a plurality of non-overlapping unit time windows according to the step length. Then, the fast Fourier transform is applied to process the data of each unit time window to obtain the frequency domain signal of each unit time window. Next, the short-time Fourier transform is applied to process each unit time window to obtain the time domain signal of each unit time window. The distribution of the frequency domain signal and the time domain signal of each unit time window is combined to obtain the time-frequency energy distribution of each unit time window.

[0060] In the feature extraction unit, the frequency feature of each unit time window is determined according to the time-frequency energy distribution of each unit time window, including the maximum value, bandwidth, variance, and energy, etc. The frequency feature of a certain unit time window is taken as the frequency feature of each time node in the unit time window. The time domain feature of each unit time window is determined by the energy fluctuation degree in the time-frequency energy distribution of each unit time window, including kurtosis, skewness, and time zero-crossing rate, etc. The time domain feature of a certain unit time window is taken as the time domain feature of each time node in the unit time window.

[0061] In one specific embodiment, the weight assignment module comprises:

[0062] a matching unit configured to calculate a matching degree between a parameter type name of a state feature of each type and an abnormal type name of each abnormal parameter type in a preset abnormal weight table through a text semantic matching technology;

[0063] an assignment unit configured to determine whether the matching degree between each parameter type name and each abnormal type name is greater than a matching degree threshold value, and if the matching degree between any parameter type name and any abnormal type name is greater than the matching degree threshold value, it is determined that the parameter type name and the abnormal type name match successfully, and the abnormal parameter type corresponding to the abnormal type name and the initial weight thereof are taken as the matching abnormal parameter type of the state feature corresponding to the parameter type name and the initial weight thereof.

[0064] Specifically, in the matching unit, the text semantic matching technology includes a bag-of-words model, a TF-IDF model, and a cosine similarity, etc., and in the embodiment of the present application, the cosine similarity is preferably used to calculate the matching degree. For example, taking Table 1 as an example, if the state feature of a certain type is a temperature type, i.e., the parameter type name is “temperature”, the matching degree between “temperature” and each abnormal type name in the preset abnormal weight table, such as “tension abnormality”, “high temperature abnormality”, etc., is calculated.

[0065] In the assignment unit, the matching degree threshold value is a threshold value obtained according to an empirical value. Still taking the above example as an example, if the matching degree between “temperature” and “high temperature abnormality” is greater than the matching degree threshold value, it is determined that “temperature” and “high temperature abnormality” match successfully, and the high temperature abnormal type corresponding to “high temperature abnormality” and the initial weight thereof are taken as the matching abnormal parameter type of the state feature corresponding to “temperature” and the initial weight thereof.

[0066] In one specific embodiment, the matching unit comprises:

[0067] a vector splitting sub-unit configured to convert the parameter type name and the abnormal type name into a state sentence vector and an abnormal sentence vector, respectively, according to a language vector model;

[0068] a matching degree calculation sub-unit configured to calculate the matching degree between each state sentence vector and each abnormal sentence vector.

[0069] Specifically, in the vector splitting sub-unit, the language vector model is a model for converting text into a vector, which maps the parameter type name and the abnormal type name into a high-dimensional space, so that semantically similar texts also exhibit similarity in the vector space, and then obtains the state sentence vector and the abnormal sentence vector; the types of the language vector model include Word2Vec, GloVe, and BERT, etc.

[0070] In one specific embodiment, the matching degree calculation subunit calculates the matching degree MD(a, b) between the state sentence vector a and the abnormal sentence vector b by formula (1):

[0071] (1);

[0072] In formula (1), represents the Euclidean distance, log( ) represents the logarithmic function, N( ) represents the number of total statistics function, N(a) represents the number of elements of the state sentence vector a, and N(b) represents the number of elements of the abnormal sentence vector b.

[0073] Specifically, in formula (1), represents the direction similarity of the state sentence vector a and the abnormal sentence vector b calculated based on the cosine similarity, represents the matching degree of the state sentence vector a and the abnormal sentence vector b in the vector dimension, and the matching degree between the state sentence vector a and the abnormal sentence vector b is obtained by combining the direction similarity and the matching degree in the vector dimension.

[0074] In one specific embodiment, the abnormality judgment module comprises:

[0075] A weight adjustment unit is configured to calculate a distribution index of each matching abnormal parameter type according to the state feature of the matching abnormal parameter type, and adjust the initial weight of each matching abnormal parameter type according to the distribution index of the matching abnormal parameter type to obtain the final weight of each matching abnormal parameter type.

[0076] A state value judgment unit is configured to calculate a state value of each potential running abnormality according to the state feature, the final weight of each matching abnormal parameter type, and the preset risk level of each potential running abnormality, and filter the potential running abnormality with a state value higher than an abnormal state threshold from all potential running abnormalities as all running abnormal states of the scraper at the current time.

[0077] Specifically, in the weight adjustment unit, the distribution index is used to represent the abnormal distribution trend of the state feature value of the matching abnormal parameter type.

[0078] In the state value judgment unit, when calculating the state value SV of a certain potential running abnormality according to the state feature, the final weight of each matching abnormal parameter type, and the preset risk level of the potential running abnormality, formula (3) is used:

[0079] (3);

[0080] In formula (3), Wi represents the final weight of the i-th matching abnormal parameter type, R represents the preset risk level of the potential running abnormality, wherein σi represents the variance of the frequency feature of the ith matching abnormal parameter type, and n represents the number of matching abnormal parameter types.

[0081] wherein the risk level of each potential running abnormality is accurately defined according to expert experience, and the abnormal state threshold is an operation abnormality threshold determined according to an experience value.

[0082] It should be noted that each potential running abnormality has a pre-set corresponding relationship with multiple abnormal parameter types. For example, if a potential running abnormality is a scraper failure, the corresponding abnormal parameter types are speed abnormality, load abnormality and current abnormality, and are irrelevant to temperature abnormality. Therefore, when calculating the state value of the scraper failure, the maximum weight of the temperature abnormality is set to 0.

[0083] In a specific embodiment, the weight adjusting unit comprises:

[0084] an abnormal matrix constructing sub-unit configured to construct an abnormal matrix according to the state feature of each matching abnormal parameter type;

[0085] a distribution index calculating sub-unit configured to input the abnormal matrix into a vector space, calculate the vector distance of the state feature of each matching abnormal parameter type in the abnormal matrix based on the vector space, and perform probability conversion on the vector distance to obtain the distribution index of each matching abnormal parameter type;

[0086] a maximum weight calculating sub-unit configured to superimpose the distribution index of each matching abnormal parameter type on the initial weight of each matching abnormal parameter type to obtain the active weight of each matching abnormal parameter type, and perform normalization processing on the active weight of all matching abnormal parameter types to obtain the maximum weight of each matching abnormal parameter type.

[0087] Specifically, in the abnormal matrix constructing sub-unit, each row of the abnormal matrix represents the state feature of a matching abnormal parameter type, and each column represents the feature value of a state feature.

[0088] In the distribution index calculating sub-unit, the vector space refers to a multi-dimensional space to which the abnormal matrix is mapped. In this multi-dimensional space, each matrix element of the abnormal matrix is represented by a vector. In the vector space, the matching abnormal parameter types in the abnormal matrix are more easily compared through nonlinear processing. The vector distance refers to the offset distance between the state feature of each matching abnormal parameter type and the state feature of other matching abnormal parameter types in the abnormal matrix. The greater the vector distance of the matching abnormal parameter type, the higher the possibility of the matching abnormal parameter type abnormality. When performing probability conversion, all vector distances are usually quantified by a Softmax function.

[0089] In the maximum weight calculation subunit, the normalization processing is to adjust the activation weight of all matching abnormal parameter types to a unified range (usually between 0 and 1), and the sum of the maximum weights of all matching abnormal parameter types is 1.

[0090] In a specific embodiment, the distribution index calculation subunit calculates the vector distance D of the state feature of a matching abnormal parameter type in the abnormal matrix by formula (2):

[0091] (2);

[0092] In formula (2), R represents the feature vector corresponding to the state feature of the matching abnormal parameter type, X represents the abnormal matrix, represents the Euclidean distance, T represents the matrix transpose, represents the sum state feature vector obtained by summing the feature vectors corresponding to the state features of each type in the abnormal matrix, and E represents the standard matrix.

[0093] Specifically, formula (2) calculates the vector distance by combining two parts, the first part is used to measure the energy distribution of the state feature of the matching abnormal parameter type in the abnormal matrix, and the second part is used to quantify the significance of the state feature of the matching abnormal parameter type relative to the abnormal matrix, and the standard matrix refers to a matrix whose elements are all 1.

[0094] Among them, the feature vector corresponding to the state feature of the matching abnormal parameter type is a vector obtained by splicing the state feature of the matching abnormal parameter type according to the arrangement type order of each column state feature in the abnormal matrix, for example, the arrangement order of a state feature in the frequency feature of the abnormal matrix is 0.5, 1.2, 0.8, and the arrangement order of the time domain feature is 0.7, 1.5, 1.0. The feature vector corresponding to the state feature is [0.5, 1.2, 0.8, 0.7, 1.5, 1.0].

[0095] In a specific embodiment, the pre-warning module comprises:

[0096] an abnormal level judgment unit configured to judge the abnormal level of each running abnormal state according to the state value of each running abnormal state of the scraper at the current time;

[0097] a pre-warning unit configured to pre-warn each running abnormal state according to the abnormal level of each running abnormal state.

[0098] Specifically, in the abnormality level judgment unit, the abnormality level of the running abnormal state is generally divided into three levels: severe, moderate and mild; each level has a state value range set according to experience value, for example, the state value range of the severe abnormality level is [0.8, 1], if the state value of a certain running abnormal state is 0.85, it is determined that the running abnormal state belongs to the severe abnormality level.

[0099] In the early warning unit, the early warning modes are different according to different abnormality levels, for example, the early warning mode of the mild abnormality level is to display a blue warning sign through the display screen or the device control panel; the early warning mode of the moderate abnormality level is to display an orange warning on the device screen, or to notify the operator through a telephone, a short message or the like; the early warning mode of the severe abnormality level is to issue a red alarm by the system, and to notify all the related personnel through an emergency broadcast, a short message, a telephone or the like.

[0100] According to all the above embodiments, the self-diagnosis and early warning system for the running state of the scraper proposed by the embodiments has the following beneficial effects:

[0101] Firstly, after the state acquisition module collects multiple types of state data of the scraper in a preset time period before the current time through multiple types of sensors arranged in advance on the scraper, the feature analysis module performs time-frequency domain decomposition processing on each type of state data to obtain the state feature of each type of state data, realizes real-time acquisition of multiple dimensions of state data of the scraper, and performs time-frequency domain decomposition to obtain the state features in the time domain and the frequency domain, thereby providing an accurate and real-time data basis for subsequent analysis of the running state abnormality of the scraper.

[0102] Then, the weight assignment module matches each type of state feature with a preset abnormality weight table to obtain the matching abnormal parameter type and the initial weight of each type of state feature, the abnormality judgment module adjusts the initial weight of each matching abnormal parameter type for the state feature of each matching abnormal parameter type to obtain the final weight, and calculates the state value of each potential running abnormality according to each matching abnormal parameter type, the final weight and the state feature, screens all running abnormal states of the scraper at the current time from all potential running abnormalities according to the state value of each potential running abnormality, and provides an abnormal state analysis mechanism that flexibly adjusts according to real-time state, thereby improving the accuracy and reliability of the analysis result of the running state abnormality of the scraper.

[0103] Finally, the early warning module early warns each running abnormal state according to the state value of each running abnormal state of the scraper at the current time, and can realize real-time, comprehensive and accurate diagnosis and early warning of the scraper.

[0104] The above merely describes the preferred embodiments of the present application and is not intended to limit the present application. It should be pointed out that for those skilled in the art, the specific embodiments of the present application can be further described in detail with reference to the accompanying drawings and examples without departing from the technical principles of the present application. On the premise of the above, a number of improvements and variations can also be made, and these improvements and variations should also be considered as the protection scope of the present application.

Claims

1. A self-diagnosis and early warning system for the running state of a scraper, characterized in that, The method comprises the following steps: A state acquisition module is configured to acquire multiple types of state data of the scraper conveyor within a preset time period before the current time through multiple types of sensors arranged in advance on the scraper conveyor; A feature analysis module is configured to perform time-frequency domain decomposition processing on each type of state data to obtain state features of each type of state data; A weight assignment module is configured to match each type of state feature with a preset abnormal weight table to obtain a matching abnormal parameter type and an initial weight corresponding to each type of state feature; The preset abnormal weight table stores a corresponding relationship between abnormal parameter types and initial weights; The weight assignment module comprises: A matching unit is configured to calculate a matching degree between a parameter type name of each type of state feature and an abnormal type name of each abnormal parameter type in the preset abnormal weight table through a text semantic matching technology; An assignment unit is configured to determine whether the matching degree between each parameter type name and each abnormal type name is greater than a matching degree threshold value. If the matching degree between any parameter type name and any abnormal type name is greater than the matching degree threshold value, it is determined that the parameter type name and the abnormal type name are matched successfully, and the abnormal parameter type and the initial weight corresponding to the abnormal type name are taken as the matching abnormal parameter type and the initial weight of the state feature corresponding to the parameter type name; An abnormality judgment module is configured to adjust the initial weight of each matching abnormal parameter type according to the state feature of each matching abnormal parameter type to obtain a final weight, calculate a state value of each potential running abnormality according to each matching abnormal parameter type, the final weight and the state feature, and screen all running abnormal states of the scraper conveyor at the current time from all potential running abnormalities according to the state value of each potential running abnormality; The abnormality judgment module comprises: A weight adjustment unit is configured to calculate a distribution index of each matching abnormal parameter type according to the state feature of each matching abnormal parameter type, and adjust the initial weight of each matching abnormal parameter type according to the distribution index to obtain a final weight of each matching abnormal parameter type; A state value judgment unit is configured to calculate a state value of each potential running abnormality according to each matching abnormal parameter type, the final weight, the state feature and a preset risk level of each potential running abnormality, and screen potential running abnormalities with a state value higher than an abnormal state threshold value from all potential running abnormalities as all running abnormal states of the scraper conveyor at the current time; The weight adjustment unit comprises: An abnormal matrix construction subunit is configured to construct an abnormal matrix according to the state feature of each matching abnormal parameter type; A distribution index calculation subunit is configured to input the abnormal matrix into a vector space, calculate a vector distance of the state feature of each matching abnormal parameter type in the abnormal matrix based on the vector space, and perform probability conversion on the vector distance to obtain a distribution index of each matching abnormal parameter type. The most weight calculation subunit is configured to superimpose a distribution index of each matched abnormal parameter type on an initial weight of each matched abnormal parameter type to obtain an activation weight of each matched abnormal parameter type, and normalize the activation weights of all matched abnormal parameter types to obtain a most weight of each matched abnormal parameter type; The early warning module is configured to perform early warning on each running abnormal state according to a state value of each running abnormal state of the scraper at a current time.

2. A self-diagnosis and early warning system for the running state of a scraper according to claim 1, characterized in that, The feature analysis module includes: The energy distribution acquisition unit is configured to decompose state data of each type in a preset time period into a plurality of unit time windows through a sliding time window, and acquire a time-frequency energy distribution of each unit time window according to Fourier transform; The feature extraction unit is configured to acquire a frequency feature and a time domain feature of state data of each type at each time node according to the time-frequency energy distribution of each unit time window of each type, wherein the frequency feature and the time domain feature of each type constitute a state feature of the state data of each type.

3. The self-diagnosis and early warning system for the running state of a scraper according to claim 1, characterized in that, The matching unit includes: The vector splitting subunit is configured to convert the parameter type name and the abnormal type name into a state sentence vector and an abnormal sentence vector, respectively, according to a language vector model; The matching degree calculation subunit is configured to calculate a matching degree between each state sentence vector and each abnormal sentence vector.

4. The self-diagnosis and early warning system for the running state of a scraper according to claim 3, characterized in that, When calculating the matching degree MD(a, b) between the state sentence vector a and the abnormal sentence vector b, the matching degree calculation subunit is implemented through formula (1): （1）； In Equation (1), represents the Euclidean distance, log() represents the logarithmic function, N() represents the number total statistical function, N(a) represents the number of elements of the state sentence vector a, and N(b) represents the number of elements of the abnormal sentence vector b.

5. The self-diagnosis and early warning system for the running state of a scraper according to claim 1, characterized in that, When calculating the vector distance D of the state feature of any matched abnormal parameter type in the abnormal matrix based on the vector space, the distribution index calculation subunit is implemented through formula (2): （2）； In formula (2), R represents a feature vector corresponding to a state feature of the matched abnormal parameter type, X represents an abnormal matrix, represents a Euclidean distance, T represents a matrix transpose, represents a sum state feature vector obtained by summing feature vectors corresponding to state features of each type in the abnormal matrix, and E represents a standard matrix.

6. A self-diagnosis and early warning system for the running state of a scraper according to claim 1, characterized in that, The early warning module includes: The abnormal level judgment unit is configured to judge an abnormal level of each running abnormal state according to a state value of each running abnormal state of the scraper at a current time; The early warning unit is configured to perform early warning on each running abnormal state according to the abnormal level of each running abnormal state.

Images