Knowledge and data dual-drive hydraulic support intelligent operation and maintenance method

Abstract

The invention provides a knowledge and data dual-drive hydraulic support intelligent operation and maintenance method, and relates to the technical field of coal mine equipment safety production. The method comprises the following steps: acquiring at least one current working condition data of the hydraulic support; inputting the at least one item of working condition data into a pre-trained fault prediction model to obtain a first prediction result output by the fault prediction model; determining a target reference case matched with the first prediction result based on the first prediction result and a plurality of candidate reference cases in an operation and maintenance knowledge base; and determining an operation and maintenance scheme of the hydraulic support based on the target reference case. Therefore, monitoring data in the production process and qualitative knowledge such as expert experience can be fused, intelligent operation and maintenance of the hydraulic support driven by data and knowledge are achieved, decision basis and theoretical reference are provided for operation and maintenance of the hydraulic support, and powerful guarantee is provided for safe production of coal mine equipment.

Description

Technical Field

[0001] The present application relates to the technical field of safe production of coal mine equipment, and in particular to an intelligent operation and maintenance method of hydraulic supports driven by both knowledge and data. Background Art

[0002] As one of the most core equipment in the fully mechanized mining face in underground mining, the hydraulic support plays a key role in bearing roof pressure and preventing the entry of coal gangue, providing a safe working environment for operators. However, due to the complex and changeable working conditions underground, the hydraulic support may be affected by various factors such as vibration and impact caused by roof movement, humid environment, improper operation, etc. during operation, causing the equipment to be in a high-load state for a long time, which is prone to functional failure and other faults, and may even cause safety accidents. Therefore, ensuring the normal operation of the hydraulic support, carrying out risk assessment and corresponding maintenance in a timely manner are of great significance to preventing roof accidents and ensuring safe production in mines. Summary of the invention

[0003] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0004] To this end, the first purpose of this application is to propose a knowledge and data dual-driven intelligent operation and maintenance method for hydraulic supports, so as to realize the combination of accurate monitoring and analysis of the status of hydraulic supports by real-time data and the learning and accumulation of historical working conditions by existing knowledge, effectively improve the accuracy and efficiency of fault diagnosis, and at the same time enhance the predictive ability of prior faults, thereby providing strong technical support for the intelligent operation and maintenance of hydraulic supports.

[0005] The second purpose of this application is to propose a hydraulic support intelligent operation and maintenance device driven by both knowledge and data.

[0006] To achieve the above objectives, the first embodiment of the present application proposes a knowledge and data dual-driven intelligent operation and maintenance method for hydraulic supports, including: Obtain at least one current working condition data of the hydraulic support; Inputting the at least one item of operating condition data into a pre-trained fault prediction model to obtain a first prediction result output by the fault prediction model; Determining a target reference case matching the first prediction result based on the first prediction result and a plurality of candidate reference cases in an operation and maintenance knowledge base; Based on the target reference case, an operation and maintenance plan for the hydraulic support is determined.

[0007] To achieve the above-mentioned purpose, the second embodiment of the present application proposes a knowledge and data dual-driven intelligent operation and maintenance device for hydraulic supports, comprising: A first acquisition module is used to acquire at least one current working condition data of the hydraulic support; A prediction module, used for inputting the at least one working condition data into a pre-trained fault prediction model to obtain a first prediction result output by the fault prediction model; A matching module, configured to determine a target reference case matching the first prediction result based on the first prediction result and a plurality of candidate reference cases in an operation and maintenance knowledge base; The first determination module is used to determine the operation and maintenance plan of the hydraulic support based on the target reference case.

[0008] The knowledge and data dual-driven intelligent operation and maintenance method and device of hydraulic supports provided in this application predicts the fault condition of the hydraulic supports by monitoring the working condition data of the hydraulic supports in real time, and then determines the reference case matching the fault prediction result in the operation and maintenance knowledge base, and determines the operation and maintenance plan of the hydraulic supports based on the processing plan of the reference case. In this way, the operation and maintenance method of the hydraulic supports can be determined by combining the working condition data monitored in real time and the operation and maintenance knowledge learned from the historical working conditions, effectively improving the accuracy and efficiency of fault diagnosis, while enhancing the ability to predict faults, and providing strong technical support for the intelligent operation and maintenance of hydraulic supports.

[0009] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which: Figure 1 A schematic diagram of a flow chart of a knowledge and data dual-driven intelligent operation and maintenance method for a hydraulic support provided in an embodiment of the present application; Figure 2 A flow chart of another knowledge and data dual-driven intelligent operation and maintenance method of a hydraulic support provided in an embodiment of the present application; Figure 3 A schematic diagram of a process for constructing an operation and maintenance knowledge base provided in an embodiment of the present application; Figure 4 A schematic diagram of a process for training a fault prediction model provided in an embodiment of the present application; Figure 5 A schematic diagram of the structure of a knowledge and data dual-driven intelligent operation and maintenance device for a hydraulic support provided in an embodiment of the present application. DETAILED DESCRIPTION

[0011] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0012] The following describes the knowledge and data dual-driven intelligent operation and maintenance method and device of the hydraulic support of the embodiment of the present application with reference to the accompanying drawings.

[0013] Figure 1 A flow chart of a knowledge and data dual-driven intelligent operation and maintenance method for a hydraulic support provided in an embodiment of the present application.

[0014] In the related art, the operation and maintenance method of hydraulic supports is that the staff always pays attention to the operating status of the equipment, mainly based on the individual experience of the technicians to judge whether the hydraulic supports are working within the reasonable limit, and adjust the control parameters to prevent the occurrence of failures. This operation and maintenance method relies on manual judgment and largely depends on manual experience, which is easily affected by subjective factors and leads to inaccurate results.

[0015] In addition, the current solution for determining how to operate and maintain hydraulic supports is purely data-driven, focusing on the processing of observation data of real systems. Although data-driven can better adapt to the complexity of real systems, it usually requires a large amount of data to train the model. However, for hydraulic supports, it is often difficult to obtain a large amount of valid data. In addition, data-driven models are not interpretable enough and are highly dependent on data quality.

[0016] To address this problem, an embodiment of the present application provides a knowledge- and data-driven intelligent operation and maintenance method for hydraulic supports, which combines the monitoring data during the production process with qualitative knowledge such as expert experience accumulated during the degradation process, effectively compensating for the problems of less fault data and poor model interpretability, and providing a more reliable method for fault prediction and maintenance of hydraulic supports.

[0017] like Figure 1 As shown, the knowledge and data dual-driven intelligent operation and maintenance method of hydraulic support includes the following steps: Step 101, obtaining at least one current working condition data of the hydraulic support.

[0018] Among them, the operating condition data may include at least one of the following: the vibration frequency and amplitude of the hydraulic support, the pressure borne by the hydraulic support, the inclination change of the hydraulic support, the displacement of the hydraulic support, the temperature and humidity of the environment in which the hydraulic support is located, the voltage and current at the key positions of the hydraulic support, and the load power change of the hydraulic support, etc.

[0019] In the embodiment of the present application, tools such as sensors can be used to monitor the working condition of the hydraulic support in real time to obtain at least one current working condition data.

[0020] Step 102: input at least one item of operating condition data into a pre-trained fault prediction model to obtain a first prediction result output by the fault prediction model.

[0021] The fault prediction model can be a bidirectional long short-term memory network (Bidirectional Long Short-Term Memory, Bi-LSTM) architecture, which is composed of an input layer, a bidirectional long short-term memory layer (Bi-LSTM), a fully connected layer, and an output layer. The parameter configuration of the fault prediction model can be determined in combination with expert experience.

[0022] In the embodiment of the present application, the constructed fault prediction model can be trained first. The trained fault prediction model can realize more accurate abnormality recognition of the working condition data by inputting the working condition data monitored in real time, and predict and evaluate the fault risk of the hydraulic support.

[0023] In an embodiment of the present application, the fault prediction model can extract the timing characteristics of at least one input working condition data, and output the prediction result corresponding to the current working condition of the hydraulic support, and then the occurrence of the hydraulic support fault can be predicted based on the prediction result. Specifically, the input layer in the fault prediction model receives the working condition data input into the model, and then the working condition data can be passed to the Bi-LSTM layer, which captures the bidirectional dependencies in the working condition data and fully extracts and utilizes the timing characteristics of the sequence information. The output of the Bi-LSTM is then further passed to the fully connected layer, and the hidden state (HiddenState) is deeply processed and converted to extract features that have a significant impact on the prediction results. Finally, the output of the fully connected layer is passed to the output layer to generate the final prediction result.

[0024] It should be noted that in the embodiment of the present application, all modeling indicators of the fault prediction model of the hydraulic support are also set based on expert knowledge, that is, all types of working conditions used by the model to predict faults, and a reasonable operating range is set to improve the fault prediction model's understanding of the sensor data. Therefore, after the real-time monitored working condition data is input into the prediction fault model, the working condition data can be standardized before being used to generate the prediction result.

[0025] Optionally, an operating range associated with an indicator corresponding to each item of operating condition data may be determined first.

[0026] In the embodiment of the present application, the modeling index corresponding to each working condition data can be determined according to the type of each working condition data, such as the vibration frequency and amplitude of the hydraulic support, the pressure borne by the hydraulic support, and the change in the inclination angle of the hydraulic support, etc. The value range or restriction condition corresponding to each indicator can be determined based on the historical fluctuation range, trend and expert knowledge (such as industry standards and specifications), so as to obtain the operating range associated with each modeling indicator during model prediction.

[0027] Then, based on the operating range, the normalized operating condition data is determined. After that, the first prediction result output by the fault prediction model can be obtained according to the normalized operating condition data and the trained fault prediction model.

[0028] In the embodiment of the present application, according to the operating range associated with each indicator, the calculation process of standardizing the operating condition data corresponding to the indicator can be shown as the following formula (1).

[0029] (1) in, is the working condition data, The working condition data after standardization processing is is the minimum value in the operating range, is the maximum value of the operating range.

[0030] Step 103: Determine a target reference case that matches the first prediction result based on the first prediction result and a plurality of candidate reference cases in the operation and maintenance knowledge base.

[0031] In this application, the prediction results may include information such as faulty components, fault types, and fault probabilities. Candidate reference cases in the operation and maintenance knowledge base can be obtained by extracting entities, determining the relationship between entities, and combining them from professional knowledge, on-site technical information, historical operation and maintenance experience, and cases. They are qualitative knowledge, and each candidate reference case may include multiple parts such as fault location (component), fault type, fault cause, and treatment method.

[0032] Therefore, in the embodiment of the present application, the first prediction result output by the hydraulic support fault prediction model can be aligned with the candidate reference cases in the operation and maintenance knowledge base, and one or more target reference cases matching the first prediction result can be found in the operation and maintenance knowledge base by interpreting the result information such as the fault type and probability output by the model. For example, the fault type in the prediction result can be matched with the fault type of the candidate reference case, and the candidate reference case with the same fault type can be determined as the target reference case.

[0033] For example, candidate reference cases in the operation and maintenance knowledge base may include, but are not limited to, the various fault cases shown in Table 1 below.

[0034] Table 1

[0035] It should be noted that in the present application, the entity and relationship information corresponding to the prediction result can also be identified based on semantic decomposition and fuzzy matching, and the knowledge computing capability of the knowledge graph can be relied upon to search for target reference cases matching the entity and relationship information in the knowledge graph corresponding to the operation and maintenance knowledge base.

[0036] Step 104, determining an operation and maintenance plan for the hydraulic support based on the target reference case.

[0037] In the embodiment of the present application, since the fault condition of the target reference case is consistent with the fault prediction result of the current hydraulic support, the fault cause and treatment method of the target reference case can provide a reference for the operation and maintenance of the current hydraulic support, and the maintenance experience and solution of the target reference case can be determined as the current operation and maintenance plan of the hydraulic support. Therefore, after determining the target reference case, an operation and maintenance plan of the hydraulic support can be generated based on the fault cause and treatment method in the target reference case.

[0038] In this embodiment, the fault condition of the hydraulic support is predicted by monitoring the working condition data of the hydraulic support in real time, and then a reference case matching the fault prediction result is determined in the operation and maintenance knowledge base, and the operation and maintenance plan of the hydraulic support is determined based on the processing plan of the reference case. In this way, the operation and maintenance method of the hydraulic support can be determined by combining the working condition data monitored in real time and the operation and maintenance knowledge learned from the historical working conditions, effectively improving the accuracy and efficiency of fault diagnosis, while enhancing the ability to predict faults, and providing strong technical support for the intelligent operation and maintenance of the hydraulic support.

[0039] This embodiment provides another intelligent operation and maintenance method of hydraulic support driven by knowledge and data. Figure 2 A flow chart of another knowledge and data dual-driven intelligent operation and maintenance method for a hydraulic support provided in an embodiment of the present application.

[0040] like Figure 2 As shown, the knowledge and data dual-driven intelligent operation and maintenance method of hydraulic support can include the following steps: Step 201, obtaining at least one current working condition data of the hydraulic support.

[0041] Step 202: input at least one item of operating condition data into a pre-trained fault prediction model to obtain a first prediction result output by the fault prediction model.

[0042] For detailed description of the above steps 201 and 202, please refer to other embodiments of the present application and will not be repeated here.

[0043] Step 203: determine the first fault type with the highest corresponding probability in the first prediction result.

[0044] In the embodiment of the present application, the first prediction result may include probability distributions of multiple fault types, and the first prediction result may be shown in the following formula (2).

[0045] (2) in, is the i-th fault type, is the probability corresponding to the i-th fault type.

[0046] The following formula (3) can be used to determine the first fault type with the highest probability in the first prediction result.

[0047] (3) in, Indicates the first fault type.

[0048] Step 204 : determining a target reference case according to the similarity between the first fault type and the second fault type corresponding to each candidate reference case in the operation and maintenance knowledge base.

[0049] In the embodiment of the present application, after determining the first fault type corresponding to the first prediction result, the similarity between the first fault type and the second fault type corresponding to each candidate reference case in the operation and maintenance knowledge base can be calculated. For example, the first fault type and the second fault type can be converted into vectors, and the cosine similarity between the two vectors can be calculated, or other methods can be used to calculate the similarity, etc. Then, the candidate reference case corresponding to the second fault type with the highest similarity can be determined as the target reference case, or all candidate reference cases with similarities greater than the similarity threshold can be determined as the target reference case.

[0050] In this embodiment, by searching for reference cases that are strongly correlated with the fault type based on the fault type, and then generating an operation and maintenance plan based on the reference case, it helps to reduce blindness and uncertainty in the fault diagnosis process, improve the quality and efficiency of operation and maintenance work, and reduce operation and maintenance costs.

[0051] It should be noted that, in the present application, the operation and maintenance knowledge base can be a graph database, which can implement the operation of screening out a target reference case from multiple candidate reference cases through the knowledge computing capability of the knowledge graph.

[0052] Optionally, based on the knowledge graph corresponding to the operation and maintenance knowledge base, at least one first entity matching the first prediction result in the knowledge graph can be determined.

[0053] It should be noted that the knowledge graph corresponding to the operation and maintenance knowledge base is obtained by extracting entities and relationships from the data used to construct the operation and maintenance knowledge base, and then drawing entities as nodes and relationships as edges. In the knowledge graph, different nodes represent different entities, which may include equipment components, fault phenomena, causes, etc. The edges between nodes represent the relationship between two entities, such as what fault phenomena exist in the components, what may be the cause of the fault phenomena, etc.

[0054] In an embodiment of the present application, the first prediction result may include components that may fail and possible failure phenomena predicted after analyzing the working conditions of the current hydraulic support, so all nodes can be traversed in the knowledge graph to determine at least one first entity that matches the first prediction result.

[0055] Then, according to at least one first entity and the relationship corresponding to the first entity, multiple candidate reference cases in the operation and maintenance knowledge can be traversed to determine a target reference case matching the first prediction result. In an embodiment of the present application, a corresponding entity node can be determined in the knowledge graph based on at least one entity in the prediction result, and then the relationship between the entities can be determined by the edges connecting the corresponding entity nodes in the knowledge graph, so that a target reference case matching the prediction result can be determined. By using the knowledge graph to search and recommend reference cases for determining operation and maintenance solutions, the efficiency of determining reference cases can be improved, thereby improving the efficiency of fault operation and maintenance.

[0056] Step 205: Determine an operation and maintenance plan for the hydraulic support based on the target reference case.

[0057] It should be noted that since the failure of the same component may be caused by multiple reasons and the same failure type may match multiple reference cases, there may be multiple target reference cases for determining the operation and maintenance plan of the hydraulic support. Therefore, when determining the operation and maintenance plan of the hydraulic support, the reference degree of each target reference case for operation and maintenance can also be considered.

[0058] Optionally, when there are multiple target reference cases, the probability and / or correlation degree corresponding to each target reference case may be determined according to the first prediction result.

[0059] In an embodiment of the present application, the probability of the fault type corresponding to each target reference case in the first prediction result can be determined, and it can be used as the probability corresponding to the target reference case. In the case where the target reference cases are all matched based on the fault type with the highest probability in the first prediction result, the probability corresponding to each target reference case is the same, then the similarity between the fault type corresponding to each target reference case and the fault type with the highest probability in the first prediction result can be determined as the degree of association corresponding to the target reference case.

[0060] Then, a weight value corresponding to the target reference case may be determined based on the probability and / or the degree of association.

[0061] In the embodiment of the present application, the higher the probability and the degree of association, the greater the weight value corresponding to the target reference case, which can indicate that the maintenance plan and troubleshooting method corresponding to the target reference case are more valuable for reference to the current status of the hydraulic support.

[0062] Afterwards, the operation and maintenance plan of the hydraulic support can be determined based on the target reference case and the corresponding weight value.

[0063] In the embodiment of the present application, the higher the probability and correlation degree corresponding to any fault type, the higher the weight value corresponding to the target reference case determined by the fault type. When determining the operation and maintenance plan of the hydraulic support based on multiple target reference cases, the processing plan of each target reference case can be sorted in descending order according to the weight value. This can improve the reliability of the operation and maintenance plan and facilitate the operation and maintenance personnel to quickly obtain the best fault prevention and processing methods.

[0064] This embodiment provides a process for constructing an operation and maintenance knowledge base for realizing the intelligent operation and maintenance method of a hydraulic support driven by both knowledge and data in the above embodiment. Figure 3 A flowchart of constructing an operation and maintenance knowledge base provided in an embodiment of the present application.

[0065] like Figure 3 As shown, the construction process of the operation and maintenance knowledge base may include the following steps: Step 301, obtaining text data related to the hydraulic support.

[0066] In the embodiment of the present application, text data related to the hydraulic support can be obtained from professional books, on-site technical data (including equipment operation records, fault handling reports, maintenance plans, etc.), maintenance experience (including fault diagnosis, maintenance skills, equipment maintenance, etc.) and production practice cases (including equipment failure signs, causes, and treatment measures).

[0067] Step 302: extract features from the text data to obtain vocabulary features corresponding to the text data and relative position features corresponding to the vocabulary features.

[0068] Among them, the vocabulary feature refers to the keywords in the text data, which can include the word part-of-speech tag, word frequency and other features. The relative position feature is the relative position relationship of the vocabulary in the text, and the relative position relationship can reflect the semantic relationship between words.

[0069] In the embodiment of the present application, the lexical features corresponding to the text data and the relative position features corresponding to the lexical features can be extracted through text cleaning, Chinese word segmentation, part-of-speech tagging, etc.

[0070] Step 303: Encode the vocabulary features and relative position features to obtain a vector corresponding to the text data.

[0071] In the embodiment of the present application, the vocabulary features and relative position features can be input into a bidirectional long short-term memory network (Bi-LSTM) for encoding. The bidirectional long short-term memory network converts the input feature sequence into a high-dimensional vector representation, such as , T is a positive integer, so we can get the vector corresponding to the text data. The position of each word in the vector space reflects its semantic relationship with other words.

[0072] Step 304: Determine a relationship between at least one second entity and the second entity based on the vector.

[0073] In the embodiment of the present application, after determining the vector corresponding to the text data, the forward hidden state can be calculated first. and the backward hidden state ,in, is the tth vector in the high-dimensional vector. Then concatenate the forward and backward hidden states to get the final hidden state .

[0074] In the embodiment of the present application, after the hidden state is obtained, a conditional random field (CRF) may be used to label the hidden state, identify at least one second entity, and implement entity extraction.

[0075] It should be noted that the CRF model is a conditional probability distribution model of another set of output random variables Y under the condition of a set of input random variables X, expressed as P(Y|X), and the formula is shown in the following formula (4).

[0076] (4) in, is the normalization factor, is the characteristic function, is the weight parameter of the feature function.

[0077] In the embodiment of the present application, the attention mechanism can be used to capture the long-distance dependency between entities and determine the relationship between the second entities. Specifically, the attention mechanism can calculate the query vector and key vector similarity and generate a weighted value , as shown in the following formula (5), to determine the dependency relationship between each word in the vector and other words, so as to obtain the relationship between the second entities.

[0078] (5) (6) Among them, in formula (6) is a vector of values, The attention calculation result is in the form of key-value pairs, reflecting the long-distance dependencies between entities.

[0079] It should be noted that after determining at least one second entity corresponding to the text data, the same entities can be fused and unified through entity alignment, entity disambiguation, etc., to eliminate the phenomenon of different expressions of the same entity that affects knowledge quality when storing multi-source knowledge.

[0080] In an embodiment of the present application, a graph neural network-based method can be used for entity alignment, as shown in the following formula (7).

[0081] (7) in, Graph neural networks take graphs as input and output a label for each node (corresponding to an entity). Each node contains information about its neighboring nodes. is the normalized adjacency matrix of the self-connected input graph, is the identity matrix, is the degree matrix, It is the first The hidden state matrix of the layer, It is The weight matrix of the layer, is the activation parameter.

[0082] In an embodiment of the present application, entity disambiguation can be achieved using a method based on context information.

[0083] Step 305: Generate a knowledge graph based on the second entity and the relationship between the second entity.

[0084] In an embodiment of the present application, at least one second entity can be used as a node in a knowledge graph, and based on the relationship between the second entities, the edge connecting the node corresponding to the second entity in the knowledge graph is determined, so that a knowledge graph can be generated.

[0085] It should be noted that after generating the knowledge graph, the incremental update method can be used to add new data or knowledge to update the knowledge graph based on the data of the original knowledge graph. For example, the incremental update algorithm can use the differential update method, and only the changed part needs to be updated. Then the updated knowledge graph satisfies the relationship with the knowledge graph before the update. G ' = G ∪Δ G , where Δ G New or modified data.

[0086] Optionally, after determining the target reference case that matches the first prediction result, the working condition data corresponding to the first prediction result can be associated with the target reference case and stored in the operation and maintenance knowledge base. The knowledge graph is incrementally updated to further improve and optimize the intelligent operation and maintenance knowledge base of the hydraulic support, realize the effective integration of quantitative and qualitative knowledge, and provide more accurate and efficient fault handling solutions for operation and maintenance personnel.

[0087] Step 306: Generate at least one reference case based on the text data, the second entities corresponding to the text data, and the relationship between the second entities.

[0088] In an embodiment of the present application, based on the connection relationship between each second entity and other entities in the knowledge graph, it is possible to determine from a component in the text data the possible fault type of the component, the possible causes of the fault type, and the method for handling the fault, etc., to generate at least one reference case. The reference case can be shown in Table 1 in the above embodiment.

[0089] It should be noted that after each reference case is generated, the reference case can be associated with the node corresponding to the entity in the knowledge graph used to generate the reference case, so that during real-time prediction, the entity and relationship information can be queried in the knowledge graph through semantic decomposition and fuzzy matching to identify the reference case.

[0090] Step 307: construct an operation and maintenance knowledge base based on the knowledge graph and at least one reference case.

[0091] In this embodiment, by performing feature extraction on text data related to the operation and maintenance knowledge of hydraulic supports, entities in the text and the relationships between entities are obtained, and then a knowledge graph is constructed based on the entities and the relationships between entities. This can improve the orderliness of the data in the operation and maintenance knowledge base and provide conditions for improving the diagnosis of hydraulic support faults, quickly locating the causes of faults, and obtaining solutions.

[0092] This embodiment provides a training process for a fault prediction model of a hydraulic support intelligent operation and maintenance method driven by knowledge and data in the above embodiment. Figure 4A schematic diagram of a process for training a fault prediction model provided in an embodiment of the present application.

[0093] like Figure 4 As shown, the training process of the fault prediction model may include the following steps: Step 401, obtaining a training data set.

[0094] The training data set includes historical operating data corresponding to at least one indicator and reference results corresponding to the historical operating data.

[0095] It should be noted that the historical operating data in the training data set should correspond to the modeling indicators of the fault prediction model. The modeling indicators include but are not limited to the vibration data of the hydraulic support, the pressure it bears, the inclination change state of the support, the displacement state, the voltage and current at key positions, etc.

[0096] In an embodiment of the present application, historical operating condition data corresponding to at least one indicator and actual fault conditions (ie, reference results) corresponding to the historical operating condition data may be collected as a training data set.

[0097] It should be noted that after obtaining the historical operating condition data, the operating condition data can be preprocessed to ensure the accuracy and consistency of the operating condition data used for prediction. For example, the operating condition data can be aligned to ensure that the timestamps of all data are consistent; missing values ​​can also be filled using methods such as k-nearest neighbors or random forest filling, and statistical methods (such as the 3σ principle) can be used to check and process outliers and delete duplicate values.

[0098] Step 402: input the historical operating condition data into the initial fault prediction model to obtain a second prediction result output by the initial fault prediction model.

[0099] In an embodiment of the present application, the input layer in the initial fault prediction model can receive historical operating condition data, and then pass the historical operating condition data to the Bi-LSTM layer to capture the bidirectional dependencies in the sequence data and fully extract and utilize the timing characteristics of the sequence information. The detailed calculation process can refer to the above-mentioned embodiment of the present application using Bi-LSTM to extract entities and relationships.

[0100] In the embodiment of the present application, the output of the Bi-LSTM can be further passed to the fully connected layer to perform in-depth processing and transformation on the hidden state, and extract features that have a significant impact on the prediction results. The calculation formula is shown in the following formula (8). Finally, the output of the fully connected layer is passed to the output layer to generate the final prediction result. The calculation formula is shown in the following formula (9).

[0101] (8) in, is the output of the fully connected layer, is the weight matrix, is the hidden state, is the bias term.

[0102] (9) in, is the final prediction result. is the weight matrix of the output layer, is the bias term.

[0103] It should be noted that expert calibration can be introduced into the initial fault prediction model to set a reasonable operating range for all modeling indicators of the hydraulic support fault prediction model based on expert knowledge, thereby improving the model's understanding of sensor data.

[0104] Step 403: determine the error between the reference result and the second prediction result.

[0105] In the embodiment of the present application, three performance indicators, namely, root mean square error (RSME), mean absolute percentage error (MAPE), and mean absolute error (MAE), can be used to respectively calculate the error between the reference result and the second prediction result. The formulas for calculating the error using the three performance indicators are shown in the following formula (10).

[0106]

[0107]

[0108] (10) in, is the second prediction result obtained for the i-th historical operating condition data, is the reference result corresponding to the i-th historical operating condition data.

[0109] Step 404: when the error is greater than the error threshold, determine a correction gradient.

[0110] The error threshold may be set based on experience, model accuracy requirements, etc., and this application does not impose any limitation on this.

[0111] In an embodiment of the present application, the error values ​​obtained by RSME, MAPE and MAE respectively can be averaged, and then the averaged error value can be compared with the error threshold. When the error is greater than the error threshold, it can be determined that the prediction accuracy of the current fault prediction model is poor, and it is necessary to perform a correction gradient update on the model parameters. For example, the gradient of the loss function with respect to the model parameters can be calculated through the back propagation algorithm, which indicates the speed and direction of the error between the reference result and the prediction result as the model parameters change, so that the correction gradient for adjusting the model parameters can be determined.

[0112] Step 405, based on the modified gradient, the initial fault prediction model is updated, and based on the updated fault prediction model, a new second prediction result is returned to obtain until the error is less than or equal to the error threshold, thereby obtaining a trained fault prediction model.

[0113] In the embodiment of the present application, the correction gradient can be used to adjust the parameter configuration of the fully connected layer of the initial fault prediction model to optimize the prediction performance of the model and complete the update of the model. The update of the fault prediction model is iterated multiple times. After each update, the historical operating condition data is input into the updated model to obtain a new prediction result. The error between the prediction result and the reference result is calculated until the error is less than or equal to the error threshold. It can be determined that the model prediction performance is good at this time, and the update iteration of the model can be stopped to obtain a trained fault prediction model. It should be noted that in the embodiments of the present application, the fault prediction model can be updated and iterated regularly based on the continuously accumulated operating condition data to ensure that the model can adapt to the latest operating condition requirements and has higher robustness.

[0114] In this embodiment, by training the fault prediction model based on the error between the reference result and the prediction result, the prediction performance of the fault prediction model can be improved, thereby providing conditions for improving the accuracy of the operation and maintenance solution.

[0115] In order to implement the above-mentioned embodiments, the present application also proposes a hydraulic support intelligent operation and maintenance device driven by both knowledge and data.

[0116] Figure 5 A schematic diagram of the structure of a knowledge and data dual-driven intelligent operation and maintenance device for a hydraulic support provided in an embodiment of the present application.

[0117] like Figure 5 As shown, the knowledge and data dual-driven hydraulic support intelligent operation and maintenance device 500 includes: The first acquisition module 501 is used to acquire at least one current working condition data of the hydraulic support; A prediction module 502, used to input at least one working condition data into a pre-trained fault prediction model to obtain a first prediction result output by the fault prediction model; A matching module 503, configured to determine a target reference case matching the first prediction result based on the first prediction result and a plurality of candidate reference cases in the operation and maintenance knowledge base; The first determination module 504 is used to determine the operation and maintenance plan of the hydraulic support based on the target reference case. Furthermore, in a possible implementation of the embodiment of the present application, the matching module 503 may be specifically configured to: Determine a first fault type having the highest corresponding probability in the first prediction result; A target reference case is determined according to a similarity between the first fault type and a second fault type corresponding to each candidate reference case in the operation and maintenance knowledge base.

[0118] Furthermore, in a possible implementation of the embodiment of the present application, the matching module 503 may be specifically configured to: Based on the knowledge graph corresponding to the operation and maintenance knowledge base, determine at least one first entity in the knowledge graph that matches the first prediction result, wherein the first entity is a node in the knowledge graph, and the edge in the knowledge graph represents the relationship between entities; According to at least one first entity and the relationship corresponding to the first entity, multiple candidate reference cases in the operation and maintenance knowledge base are traversed to determine a target reference case that matches the first prediction result.

[0119] Furthermore, in a possible implementation of the embodiment of the present application, the prediction module 502 may be specifically used to: Determine the operating range associated with the indicator corresponding to each operating condition data; Based on the operating range, determine the operating condition data after standardization; According to the standardized operating condition data and the fault prediction model, a first prediction result output by the fault prediction model is obtained.

[0120] Furthermore, in a possible implementation of the embodiment of the present application, the knowledge and data dual-driven intelligent operation and maintenance device 500 for hydraulic support may further include: The second acquisition module is used to acquire text data related to the hydraulic support; A feature extraction module is used to extract features from text data to obtain lexical features corresponding to the text data and relative position features corresponding to the lexical features; The encoding module is used to encode the vocabulary features and relative position features to obtain the vector corresponding to the text data; A second determination module, configured to determine a relationship between at least one second entity and a second entity based on the vector; The first generating module is used to generate a knowledge graph according to the second entity and the relationship between the second entity.

[0121] The second generating module is used to generate at least one reference case according to the text data, the second entities corresponding to the text data and the relationship between the second entities.

[0122] The third generation module is used to build an operation and maintenance knowledge base based on the knowledge graph and at least one reference case.

[0123] Furthermore, in a possible implementation of the embodiment of the present application, the knowledge and data dual-driven intelligent operation and maintenance device 500 for hydraulic support may further include: The storage module is used to associate the operating condition data corresponding to the first prediction result with the target reference case after determining the target reference case matching the first prediction result, and store it in the operation and maintenance knowledge base.

[0124] Furthermore, in a possible implementation of the embodiment of the present application, the knowledge and data dual-driven intelligent operation and maintenance device 500 for hydraulic support may further include: A third acquisition module is used to acquire a training data set, wherein the training data set includes historical operating condition data corresponding to at least one indicator and a reference result corresponding to the historical operating condition data; A fourth acquisition module, used for inputting the historical operating condition data into the initial fault prediction model to obtain a second prediction result output by the initial fault prediction model; A third determination module, used to determine the error between the reference result and the second prediction result; A fourth determination module, used to determine a correction gradient when the error is greater than an error threshold; The updating module is used to update the initial fault prediction model based on the modified gradient, and return a new second prediction result based on the updated fault prediction model until the error is less than or equal to the error threshold, thereby obtaining a trained fault prediction model.

[0125] Further, in a possible implementation of the embodiment of the present application, the first determining module 504 may be specifically configured to: In the case where there are multiple target reference cases, determining the probability and / or degree of association corresponding to each target reference case according to the first prediction result; Determine the weight value corresponding to the target reference case based on the probability and / or the degree of association; Based on the target reference case and the corresponding weight value, the operation and maintenance plan of the hydraulic support is determined.

[0126] It should be noted that the above explanation of the embodiment of the hydraulic support intelligent operation and maintenance method driven by both knowledge and data is also applicable to the hydraulic support intelligent operation and maintenance device driven by both knowledge and data of this embodiment, and will not be repeated here.

[0127] In the embodiment of the present application, the fault condition of the hydraulic support is predicted by monitoring the working condition data of the hydraulic support in real time, and then a reference case matching the fault prediction result is determined in the operation and maintenance knowledge base, and the operation and maintenance plan of the hydraulic support is determined based on the processing plan of the reference case. In this way, the operation and maintenance method of the hydraulic support can be determined by combining the working condition data monitored in real time and the operation and maintenance knowledge learned from the historical working conditions, effectively improving the accuracy and efficiency of fault diagnosis, while enhancing the ability to predict faults, and providing strong technical support for the intelligent operation and maintenance of the hydraulic support.

[0128] The collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in this application are in compliance with relevant laws and regulations and do not violate public order and good morals.

[0129] It should be noted that personal information from users should be collected for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. In addition, such collection / sharing should be carried out after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign the agreement / authorization including authorization of relevant user information before the user uses the function. In addition, any necessary steps should be taken to protect and safeguard access to such personal information data and ensure that others who have access to personal information data comply with its privacy policy and procedures.

[0130] The present application is expected to provide an implementation scheme for users to selectively block the use or access of personal information data. That is, the present disclosure is expected to provide hardware and / or software to prevent or block access to such personal information data. Once the personal information data is no longer needed, the risk can be minimized by limiting data collection and deleting the data. In addition, when applicable, such personal information is de-identified to protect the privacy of the user.

[0131] In the description of the aforementioned embodiments, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they contradict each other.

[0132] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0133] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0134] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0135] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0136] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0137] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0138] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A knowledge and data dual-driven intelligent operation and maintenance method for hydraulic supports, characterized in that: include: Obtain at least one current working condition data of the hydraulic support; Inputting the at least one item of operating condition data into a pre-trained fault prediction model to obtain a first prediction result output by the fault prediction model; Determining a target reference case matching the first prediction result based on the first prediction result and a plurality of candidate reference cases in an operation and maintenance knowledge base; Based on the target reference case, an operation and maintenance plan for the hydraulic support is determined.

2. The method according to claim 1, characterized in that The determining, based on the first prediction result and a plurality of candidate reference cases in the operation and maintenance knowledge base, a target reference case matching the first prediction result includes: Determine a first fault type having the highest corresponding probability in the first prediction result; A target reference case is determined according to a similarity between the first fault type and a second fault type corresponding to each candidate reference case in the operation and maintenance knowledge base.

3. The method according to claim 1, characterized in that The determining, based on the first prediction result and a plurality of candidate reference cases in the operation and maintenance knowledge base, a target reference case matching the first prediction result includes: Based on the knowledge graph corresponding to the operation and maintenance knowledge base, determine at least one first entity in the knowledge graph that matches the first prediction result, wherein the first entity is a node in the knowledge graph, and the edge in the knowledge graph represents the relationship between entities; According to the at least one first entity and the relationship corresponding to the first entity, multiple candidate reference cases in the operation and maintenance knowledge base are traversed to determine a target reference case that matches the first prediction result.

4. The method according to claim 1, characterized in that The step of inputting the at least one item of operating condition data into a pre-trained fault prediction model to obtain a first prediction result output by the fault prediction model includes: Determine the operating range associated with the indicator corresponding to each operating condition data; Based on the operating range, determining the operating condition data after the standardized processing; According to the standardized operating condition data and the fault prediction model, a first prediction result output by the fault prediction model is obtained.

5. The method according to any one of claims 1 to 3, characterized in that: The construction process of the operation and maintenance knowledge base includes: Get text data related to hydraulic support; Extracting features from the text data to obtain vocabulary features corresponding to the text data and relative position features corresponding to the vocabulary features; Encoding the vocabulary features and the relative position features to obtain a vector corresponding to the text data; determining, based on the vector, a relationship between at least one second entity and the second entity; Generate a knowledge graph according to the relationship between the second entity and the second entity; generating at least one reference case according to the text data, the second entity corresponding to the text data, and the relationship between the second entities; An operation and maintenance knowledge base is constructed based on the knowledge graph and the at least one reference case.

6. The method according to claim 5, characterized in that Also includes: After determining a target reference case that matches the first prediction result, the operating condition data corresponding to the first prediction result is associated with the target reference case and stored in the operation and maintenance knowledge base.

7. The method according to claim 1, characterized in that The training process of the fault prediction model includes: Acquire a training data set, wherein the training data set includes historical operating condition data corresponding to at least one indicator and a reference result corresponding to the historical operating condition data; Inputting the historical operating condition data into an initial fault prediction model to obtain a second prediction result output by the initial fault prediction model; determining an error between the reference result and the second prediction result; When the error is greater than the error threshold, determining a correction gradient; Based on the corrected gradient, the initial fault prediction model is updated, and based on the updated fault prediction model, a new second prediction result is returned to obtain until the error is less than or equal to the error threshold, thereby obtaining a trained fault prediction model.

8. The method according to claim 1, characterized in that The step of determining the operation and maintenance plan of the hydraulic support based on the target reference case includes: In the case where there are multiple target reference cases, determining the probability and / or correlation degree corresponding to each target reference case according to the first prediction result; Based on the probability and / or the degree of association, determining a weight value corresponding to the target reference case; Based on the target reference case and the corresponding weight value, an operation and maintenance plan of the hydraulic support is determined.

9. A knowledge and data dual-driven intelligent operation and maintenance device for hydraulic supports, characterized in that: include: A first acquisition module is used to acquire at least one current working condition data of the hydraulic support; A prediction module, used for inputting the at least one item of operating condition data into a pre-trained fault prediction model to obtain a first prediction result output by the fault prediction model; A matching module, configured to determine a target reference case matching the first prediction result based on the first prediction result and a plurality of candidate reference cases in an operation and maintenance knowledge base; The first determination module is used to determine the operation and maintenance plan of the hydraulic support based on the target reference case.

Images