Loader multi-sensor data fusion fault diagnosis and early warning method

Abstract

The invention relates to the technical field of loaders, in particular to a loader multi-sensor data fusion fault diagnosis and early warning method, which comprises the following steps: S1, multi-source data acquisition and preprocessing: acquiring running state data of a loader in real time; s2, multi-modal data fusion: carrying out multi-modal data fusion on the preprocessed operation state data to form comprehensive data; s3, energy flow fault propagation prediction: analyzing energy flow distribution of each component in real time, and identifying a potential fault component and a propagation path thereof; and S4, fault assessment and graded early warning: comprehensively assessing the fault risk of the loader, and generating a graded early warning signal according to a set multi-stage early warning rule. According to the invention, the operation efficiency and safety of the loader are improved, the distribution of maintenance resources is optimized, the operation cost is reduced, the overall reliability of the loader is further improved, and the service life of the loader is further prolonged.

Description

Technical Field

[0001] The present invention relates to the technical field of loaders, and in particular to a fault diagnosis and early warning method of multi-sensor data fusion for loaders. Background Art

[0002] With the improvement of the automation level of mechanical equipment, loaders are increasingly used in various working environments. As a heavy-duty engineering machinery, loaders undertake key tasks such as material handling and earthwork. Their operating status directly affects the construction progress and work efficiency. In order to ensure the efficient operation and safety of loaders, equipment fault diagnosis and early warning management have become increasingly important. By real-time monitoring of the loader's operating status and timely detection of potential faults, downtime can be effectively reduced, maintenance costs can be reduced, and work efficiency can be improved. Therefore, how to use advanced diagnostic technology to perform comprehensive fault diagnosis and early warning on loaders has become the key to improving their reliability and safety.

[0003] At present, loader fault diagnosis mostly relies on a single monitoring technology, such as vibration monitoring, temperature monitoring, etc. Although these traditional methods can provide some operating status information, they often cannot fully reflect the health status of the loader because they rely only on a single data source. At the same time, existing fault diagnosis methods generally have problems such as reliance on manual experience, diagnostic lag, and insufficient accuracy. Especially in fault prediction and early warning, many methods ignore the dynamic process of fault propagation and cannot identify the fault propagation path and range in a timely and accurate manner. Therefore, traditional fault diagnosis and early warning technologies are unable to cope with the complex working environment and changeable operating status of loaders, and there is a lot of room for improvement.

[0004] The purpose of the present invention is to provide a fault diagnosis and early warning method based on multi-sensor data fusion, to provide a scientific basis for maintenance decisions of loaders, to optimize the allocation of maintenance resources, to improve the reliability and safety of loaders, to significantly reduce the probability of failure and downtime, and to have important economic benefits and application value. Summary of the invention

[0005] The invention provides a fault diagnosis and early warning method for loader multi-sensor data fusion.

[0006] The fault diagnosis and early warning method of multi-sensor data fusion for loaders includes the following steps:

[0007] S1, multi-source data acquisition and preprocessing: using a variety of sensors arranged on the loader, real-time acquisition of the loader's operating status data, including vibration, temperature, pressure and current, and preprocessing of the acquired operating status data, including denoising, outlier correction, time alignment and standardization;

[0008] S2, multimodal data fusion: the pre-processed operation status data is subjected to multimodal data fusion to form comprehensive data;

[0009] S3, energy flow fault propagation prediction: Based on the power transmission path of the loader and the energy flow characteristics between the components, an energy flow fault propagation model is constructed to analyze the energy flow distribution of each component in real time and identify potential faulty components and their propagation paths, including:

[0010] S31, constructing a loader power transmission path model: constructing an energy flow fault propagation model according to the loader power transmission path and energy flow characteristics between various components (engine, hydraulic system, transmission system, steering system);

[0011] S32, Energy flow distribution analysis: Based on the energy flow fault propagation model, the energy input, output and energy loss of each component are analyzed in real time, the operating efficiency and energy loss of each component are identified, and potential faults are discovered;

[0012] S33, Fault propagation path prediction: Based on the results of energy flow distribution analysis, predict the propagation path strength, propagation rate and impact range of potential faults;

[0013] S4, fault assessment and graded warning: Based on the integrated data after fusion and combined with the results of energy flow fault propagation prediction, the failure risk of the loader is comprehensively assessed, and graded warning signals are generated according to the set multi-level warning rules, including low-risk warning, medium-risk warning and high-risk warning, and corresponding maintenance suggestions are provided for different warning signals.

[0014] Optionally, the multi-source data collection and preprocessing in S1 includes:

[0015] S11, multi-source data collection: using a variety of sensors arranged on the loader, including vibration sensors, temperature sensors, pressure sensors and current sensors, to collect the operating status data of the loader in real time;

[0016] S12, data preprocessing: preprocessing the real-time collected operation status data, specifically including:

[0017] S121, denoising: using wavelet transform method to denoise the running status data, decomposing the original signal into multi-scale wavelet coefficients, and removing high-frequency noise signals;

[0018] S122, outlier correction: detect and correct outliers by IQR (interquartile range) method;

[0019] S123, time alignment: In view of the sampling frequency differences of multiple sensors, the linear interpolation method is used to time align the operation status data and unify the timestamp;

[0020] S124, normalization processing: normalizing the running status data.

[0021] Optionally, the multimodal data fusion in S2 includes:

[0022] S21, data mapping and normalization: mapping the preprocessed operating status data to a uniform value range through a normalization method;

[0023] S22, weighted fusion of multimodal data: The normalized operating status data is subjected to a weighted fusion method to generate comprehensive data S(t).

[0024] Optionally, the energy flow fault propagation model in S31 adopts a graph neural network (GNN) model, and the graph neural network (GNN) model includes:

[0025] S311, constructing a graph structure model: Modeling the power transmission path of the loader as a weighted directed graph G = (V, E), where V is a node set representing the components of the loader (engine, hydraulic system, transmission system, steering system), E is an edge set representing the energy flow path between the components, and the weight represents the energy transmission efficiency, specifically including:

[0026] Node and edge initialization: initial feature vector for each node It is the operation status data;

[0027] Edge weight initialization: edge weight w i,j Initialized as energy transfer efficiency;

[0028] S312, dynamic node feature update: update the feature vector of each node according to the node's neighbor information and edge weight, adding the time dimension and energy flow relationship;

[0029] S313, fault propagation path calculation: Using the updated results of node features, calculate the fault propagation probability P of each edge i,j and time delay T i,j ;

[0030] S314, global fault risk assessment: Based on the final characteristics of nodes and edges, the risk scope and critical path of global fault propagation are assessed, including:

[0031] Global fault risk score: based on the fault propagation probability P of each edge i,j and time delay T i,j , calculate the global failure risk score R;

[0032] Main propagation path: Select P i,j ·T i,j The largest path is used as the main propagation path for fault propagation.

[0033] Optionally, the energy flow distribution analysis in S32 includes:

[0034] S321, Calculation of energy input, output and energy loss: For each component i, the energy input E in,i , Energy output E out,i and energy loss E loss,i Perform calculations;

[0035] S322, Evaluation of operating efficiency and energy loss: By comparing the input energy and output energy of each component, calculate its operating efficiency η i (t);

[0036] S323, fault detection and early warning: perform fault detection based on the threshold of operating efficiency and energy loss. If η i (t) is lower than the set efficiency threshold or E loss,i If the loss threshold is exceeded, the component is considered to be potentially faulty.

[0037] Optionally, the fault propagation path prediction in S33 includes:

[0038] S331, fault propagation path strength calculation: by analyzing the energy input, output and loss of each component, combined with the energy flow fault propagation model, the energy transfer relationship and fault propagation potential between components are calculated, and the fault propagation path strength is predicted;

[0039] S332, propagation rate calculation: based on the fault propagation path, analyze the propagation rate of the fault energy change per unit time from component i to component j;

[0040] S333, Impact Scope Prediction: Predict the impact scope of the fault based on the propagation path and rate, and determine the impact scope of the fault by comprehensively evaluating the fault propagation potential of each component and the overall energy transfer efficiency.

[0041] Optionally, the fault assessment and graded warning in S4 include:

[0042] S41, Fault risk assessment: Based on the comprehensive data after multimodal data fusion and the results of energy flow fault propagation prediction, the fault risk is comprehensively assessed;

[0043] S42, fault warning signal generation: setting multi-level warning rules, comparing the results of the fault risk assessment with the preset risk threshold, and generating fault warning signals of different levels, including low risk warning, medium risk warning and high risk warning;

[0044] S43, generating maintenance suggestions: automatically generating corresponding maintenance suggestions according to the generated fault warning signals of different levels.

[0045] Optionally, the fault risk assessment in S41 includes:

[0046] S411, Fault risk assessment: Based on the comprehensive data after multi-modal data fusion and the results of energy flow fault propagation prediction, the fault risk index R is calculated. fault (i, t);

[0047] S412, Comprehensively evaluate the failure risk of the loader: Use the failure risk index R of each component fault (i, t) is weighted and summed to calculate the comprehensive failure risk index R total (t).

[0048] Optionally, the generation of the fault warning signal in S42 includes:

[0049] S421, setting risk threshold: setting the risk threshold of loader failure risk, including the lower limit of the risk threshold R low and the upper risk threshold R high ;

[0050] S422, Risk level judgment and warning signal generation: Based on the fault risk assessment results and the risk threshold lower limit R low and the upper risk threshold R high By comparison, the fault risks are divided into different risk levels, and corresponding warning signals are generated according to different levels, including:

[0051] Low risk warning: When R total (t) <R low When , it indicates a low risk warning, and a low risk warning signal is generated;

[0052] Medium risk warning: When R low <R total (t) <R high When , it indicates a medium risk warning, and a medium risk warning signal is generated;

[0053] High risk warning: When R total (t)>R high When , it indicates a high-risk warning, a high-risk warning signal is generated.

[0054] Optionally, the generating of the maintenance suggestion in S43 includes:

[0055] S431, low-risk warning suggestion generation: based on the low-risk warning signal, low-risk warning suggestions are generated, including routine inspections, regular monitoring and data recording of components, and maintaining normal operation;

[0056] S432, medium risk warning suggestion generation: generating medium risk warning suggestions based on the medium risk warning signal, including inspecting the faulty parts (checking for abnormal signs or wear), arranging preventive maintenance of potential faulty parts, and inspecting the operating status of the loader and performing calibration or adjustment;

[0057] S433, high-risk warning suggestion generation: Generate high-risk warning suggestions based on high-risk warning signals, including immediate shutdown inspection and comprehensive fault diagnosis, and emergency repair or replacement of potential faulty parts.

[0058] Beneficial effects of the present invention:

[0059] The present invention, through comprehensive multi-source data collection and sophisticated preprocessing, can monitor the operating status of the loader in real time and accurately reflect the health status of the equipment. By fusing multimodal data such as vibration, temperature, pressure and current, the differences and interference between different data sources are eliminated, providing high-quality data support for subsequent fault diagnosis. In particular, combined with the results of energy flow fault propagation prediction, it can effectively identify potential faulty components and their propagation paths, and accurately predict the time and scope of the fault. This process can greatly improve the real-time and accuracy of diagnosis, provide a reliable basis for fault warning, thereby optimizing maintenance decisions and improving the safety and reliability of the system.

[0060] The present invention adopts graph neural network for modeling, so that the energy flow and fault transmission relationship between components can be accurately captured and dynamically analyzed. Through node feature updating and edge weight adjustment, the fault propagation path can be accurately calculated, including the prediction of propagation path strength, propagation rate and impact range. It can not only identify potential faults in the early stage, but also provide real-time assessment of the health status of each component. The fault risk assessment is combined with the energy flow prediction results, through weighted summation, and comprehensively considers the risk coefficient of each component to provide a quantitative assessment of the overall failure risk of the loader.

[0061] The present invention provides a flexible and efficient solution for fault management by generating multi-level warning signals and corresponding maintenance suggestions. It can automatically generate accurate maintenance suggestions for fault warning signals of different risk levels (low, medium and high risks), helping maintenance personnel to quickly identify fault points and take targeted measures in a timely manner. This not only improves the operating efficiency and safety of the loader, but also optimizes the allocation of maintenance resources, reduces operating costs, and further improves the overall reliability and service life of the loader. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0063] Figure 1 A schematic diagram of a process flow of a diagnosis and early warning method according to an embodiment of the present invention;

[0064] Figure 2 Schematic diagram of energy flow fault propagation prediction according to an embodiment of the present invention. DETAILED DESCRIPTION

[0065] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0066] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0067] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0068] like Figure 1-Figure 2 As shown, the fault diagnosis and early warning method of loader multi-sensor data fusion includes the following steps:

[0069] S1, multi-source data acquisition and preprocessing: using a variety of sensors arranged on the loader, real-time acquisition of the loader's operating status data, including vibration, temperature, pressure and current, and preprocessing of the acquired operating status data, including denoising, outlier correction, time alignment and standardization;

[0070] S2, multimodal data fusion: the pre-processed operation status data is subjected to multimodal data fusion to form comprehensive data;

[0071] S3, energy flow fault propagation prediction: Based on the power transmission path of the loader and the energy flow characteristics between the components, an energy flow fault propagation model is constructed to analyze the energy flow distribution of each component in real time and identify potential faulty components and their propagation paths, including:

[0072] S31, constructing a loader power transmission path model: constructing an energy flow fault propagation model according to the loader power transmission path and energy flow characteristics between various components (engine, hydraulic system, transmission system, steering system);

[0073] S32, Energy flow distribution analysis: Based on the energy flow fault propagation model, the energy input, output and energy loss of each component are analyzed in real time, the operating efficiency and energy loss of each component are identified, and potential faults are discovered;

[0074] S33, Fault propagation path prediction: Based on the results of energy flow distribution analysis, predict the propagation path strength, propagation rate and impact range of potential faults;

[0075] S4, fault assessment and graded warning: Based on the integrated data after fusion and the results of energy flow fault propagation prediction, the fault risk of the loader is comprehensively assessed, and graded warning signals are generated according to the set multi-level warning rules, including low-risk warning, medium-risk warning and high-risk warning. Corresponding maintenance suggestions are provided for different warning signals;

[0076] Through the above content, comprehensive monitoring and efficient diagnosis of the loader's operating status are achieved. It can not only collect and process a variety of key data in real time, but also analyze the fault propagation path through the energy flow model, accurately predict the occurrence and development of faults, and provide more efficient fault diagnosis and early warning capabilities. It can identify potential faults earlier, optimize maintenance decisions, reduce downtime, and improve the operating efficiency and reliability of the loader, which has significant economic benefits and safety assurance effects.

[0077] Multi-source data collection and preprocessing in S1 include:

[0078] S11, multi-source data collection: using a variety of sensors arranged on the loader, including vibration sensors, temperature sensors, pressure sensors and current sensors, to collect the operating status data of the loader in real time;

[0079] S12, data preprocessing: preprocessing the real-time collected operation status data, specifically including:

[0080] S121, denoising: use wavelet transform method to denoise the running status data, decompose the original signal into multi-scale wavelet coefficients, remove high-frequency noise signals, expressed as:

[0081] X ′ (t)=∑ k c k φ k (t)+∑ j≥J d j,k ψ j,k (t);

[0082] Among them, φ k (t) is the low-frequency wavelet function, ψ j,k (t) is the high frequency wavelet function, c k and d j,k are the corresponding low-frequency and high-frequency coefficients, respectively. k determines the time position, j determines the frequency scale, J sets the depth of decomposition (the highest scale), and X ′ (t) is the denoised operating status data;

[0083] S122, outlier correction: detect and correct outliers by using the IQR (interquartile range) method, calculate the first quartile (Q1) and the third quartile (Q3) of the data, and the outlier judgment conditions are:

[0084]

[0085] S123, time alignment: In view of the sampling frequency differences of multiple sensors, the linear interpolation method is used to time align the operating status data, unify the timestamps, and set the sampling interval to Δt, which is expressed as:

[0086]

[0087] Among them, t1, t2 are adjacent known time points, X(t1), X(t2) are the corresponding sensor data;

[0088] S124, standardization processing: normalize the running status data to ensure that the value ranges of different data sources are consistent to eliminate the impact of unit differences, expressed as:

[0089]

[0090] Among them, μ is the data mean, σ is the data standard deviation, and X(t) is the initial operating status data;

[0091] Through the above content, the operating status of the loader is fully captured, and the data quality and consistency are effectively improved through denoising, outlier correction, time alignment and standardization. It can not only accurately reflect the real-time working conditions of the loader, but also solve the impact of sampling frequency differences, noise interference and abnormal data between multi-source data, thereby improving the diagnostic accuracy and system robustness.

[0092] Multimodal data fusion in S2 includes:

[0093] S21, data mapping and normalization: The preprocessed running status data is mapped to a uniform numerical range through a normalization method, which is expressed as:

[0094]

[0095] Among them, X(t) represents the original running status data, X min and X max are the minimum and maximum values ​​of the original running status data respectively;

[0096] S22, weighted fusion of multimodal data: The normalized running status data is subjected to weighted fusion to generate comprehensive data S(t), which is expressed as:

[0097] S(t)=w V ·V 归一化 (t)+w T ·T 归一化 (t)+w P ·P 归一化 (t)+w I I 归一化

[0098] (t);

[0099] Among them, V 归一化 (t), T 归一化 (t), P 归一化 (t), I 归一化 (t) are the normalized results of vibration, temperature, pressure and current data, respectively, w V ,w T ,w P ,w I are the corresponding weights respectively;

[0100] Through the above content, the problems of inconsistent distribution and magnitude difference of different modal data are solved, the scientificity and accuracy of the fusion process are ensured, the ability of comprehensive data to express the fault status of the loader is improved, unnecessary links are effectively avoided, and the fusion efficiency is improved. At the same time, it provides a high-quality data foundation for subsequent fault diagnosis and early warning.

[0101] The energy flow fault propagation model in S31 adopts the graph neural network (GNN) model, which includes:

[0102] S311, constructing a graph structure model: Modeling the power transmission path of the loader as a weighted directed graph G = (V, E), where V is a node set representing the components of the loader (engine, hydraulic system, transmission system, steering system), E is an edge set representing the energy flow path between the components, and the weight represents the energy transmission efficiency, specifically including:

[0103] Node and edge initialization: initial feature vector for each node is the running status data, expressed as:

[0104]

[0105] Among them, v i ,t i 、p i 、c i are the vibration, temperature, pressure and current data of node i respectively;

[0106] Edge weight initialization: edge weight w i,j Initialized to the energy transfer efficiency, expressed as:

[0107] w i,j =η i→j ;

[0108] Among them, η i→j represents the energy transfer efficiency from node i to node j;

[0109] S312, dynamic node feature update: according to the node's neighbor information and edge weight, update the feature vector of each node, add the time dimension and energy flow relationship, expressed as:

[0110]

[0111] in, is the message sent from node j to node i, w i,j is the energy transfer efficiency obtained from the edge weights, are trainable weight matrices and biases, σ is the ReLU activation function, is the feature vector of node j in layer l;

[0112]

[0113] in, is the updated feature vector of node i at layer l+1, is the feature vector of node i in layer l, N(i) is the set of neighbor nodes of node i, Trainable weight matrix updated for node feature vector;

[0114] S313, fault propagation path calculation: Calculate the fault propagation probability P of each edge using the updated result of the node feature vector i,j and time delay T i,j , expressed as:

[0115]

[0116] Among them, P i,j is the probability of fault propagation from node i to node j, W p ,b p are trainable weight matrices and biases, respectively. is the feature vector of node i in layer L, is the feature vector of node j in layer L, Indicates the concatenation of node feature vectors and edge weights w i,j As input;

[0117]

[0118] Among them, T i,j is the time delay for the fault to propagate from node i to node j, ΔE i,j is the energy change from node i to node j, E max,j is the maximum energy capacity of node j;

[0119] S314, global fault risk assessment: Based on the final feature vectors of nodes and edges, the risk range and critical path of global fault propagation are assessed, including:

[0120] Global fault risk score: based on the fault propagation probability P of each edge i,j and time delay T i,j , calculate the global failure risk score R, expressed as:

[0121] R=∑ i∈V ∑ j∈N(i) P i,j ·T i,j ;

[0122] Where V is the set of nodes in graph G;

[0123] Main propagation path: Select P i,j ·T i,j The largest path is used as the main propagation path for fault propagation;

[0124] Through the above content, the energy interaction and fault transmission relationship between each component can be dynamically modeled. By introducing edge weights to quantitatively express the energy transfer efficiency, as well as layer-by-layer updating of node feature vectors and dynamic calculation of fault propagation probability, the fault propagation path, impact range and propagation delay can be accurately predicted. GNN can better adapt to the dynamic characteristics of complex systems, provide higher diagnostic accuracy and real-time performance, and at the same time provide powerful decision-making support for loader operation status monitoring and early warning, significantly improving the safety and reliability of the system.

[0125] Energy flow distribution analysis in S32 includes:

[0126] S321, Calculation of energy input, output and energy loss: For each component i, the energy input E in,i 、Energy output E out,i and energy loss E loss,i Calculate and express it as:

[0127]

[0128] E loss,i (t) = E in,i (t)-E out,i (t);

[0129] Among them, w k,i is the energy transfer efficiency (edge ​​weight) from component k to component i, is the state characteristic vector of component k (vibration, temperature, pressure and current), N i is the set of neighbor nodes of component i, N i ′ is the output neighbor node set of component i, t is the time variable, w i,j is the energy transfer efficiency or weight between component i and component j, is the feature vector of component j at layer L, representing the state feature vector of component j at time t;

[0130] S322, Evaluation of operating efficiency and energy loss: By comparing the input energy and output energy of each component, calculate its operating efficiency η i (t), expressed as:

[0131]

[0132] S323, fault detection and early warning: perform fault detection based on the threshold of operating efficiency and energy loss. If η i (t) is lower than the set efficiency threshold or E loss,i If the loss threshold is exceeded, the component is considered to have potential failure;

[0133] The efficiency critical value is based on historical data statistics, combined with the efficiency distribution under normal operating conditions, and is set by the lower bound of the distribution, including:

[0134] Data statistics: Extract the efficiency data η of component i under normal operating conditions from historical operation records i ;

[0135] Calculate the lower critical value η using the quantile method 临界 , expressed as:

[0136] η 临界 =Q1-m·IQR;

[0137] Where Q1 is the first quartile of operating efficiency, IQR = Q3-Q1 is the interquartile range of operating efficiency, and m is the safety factor (taken as 1.5);

[0138] The loss threshold is set based on the operating conditions and the maximum energy capacity of the component, combined with empirical rules, including:

[0139] Calculate the theoretical maximum energy loss: Based on the energy input capacity E of component i in,max,i and the transfer efficiency η 设计,i , calculate the maximum theoretical energy loss, expressed as:

[0140] E 损耗,max,i =E in,max,i ·(1-η 设计,i );

[0141] Set the loss threshold: Combine the operation redundancy factor α (1.1 to 1.2) to set the energy loss threshold, expressed as:

[0142] E 阈值,i =α·E 损耗,max,i ;

[0143] Through the above content, based on the energy flow fault propagation model, it is possible to accurately identify components with high energy loss or low efficiency and discover potential faults in a timely manner. The energy flow distribution analysis has higher dynamic adaptability and prediction capabilities, and can provide accurate basis for fault detection. At the same time, it can generate targeted early warning signals, providing important support for maintenance decisions and system optimization, thereby significantly improving the reliability and safety of loader operation.

[0144] Fault propagation path prediction in S33 includes:

[0145] S331, fault propagation path strength calculation: By analyzing the energy input, output and loss of each component, combined with the energy flow fault propagation model, the energy transfer relationship and fault propagation potential between components are calculated, and the fault propagation path strength is predicted, which is expressed as:

[0146]

[0147] Where, ΔE i,j (t) represents the energy change between component i and component j at time t, N j is the set of neighbor nodes of component j, w k,j is the energy transfer efficiency or weight from component k to component j, w i,k is the energy transfer efficiency or weight from component i to component j;

[0148]

[0149] Among them, P i,j (t) represents the propagation path strength (propagation potential) of the fault from component i to component j, ΔE i,j (t) is the energy change between component i and component j at time t, E loss,i (t) is the energy loss of component i at time t, indicating the degree of energy loss of component i under normal operation, E loss,j (t) is the energy loss of component j at time t, indicating the degree of energy loss of component j under normal operation;

[0150] S332, propagation rate calculation: Based on the fault propagation path, the time required for the fault to propagate from component i to component j is analyzed, that is, the propagation rate of the fault energy change per unit time, which is expressed as:

[0151]

[0152] Among them, v i,j (t) is the propagation rate of fault from component i to component j, ΔE i,j (t) is the energy change between component i and component j at time t, and Δt is the change per unit time;

[0153] S333, Impact Scope Prediction: Based on the propagation path and rate, the impact scope of the fault is predicted. By comprehensively evaluating the fault propagation potential of each component and the overall energy transfer efficiency, the impact scope of the fault is determined, which is expressed as:

[0154]

[0155] Among them, ImpactRange(t) is the size of the fault impact range at time t, N affected is the set of components affected by the fault, and represents the set of all components affected by the fault. loss,i (t) is the energy loss of component i at time t, which indicates the energy value lost by component i due to fault or fault propagation;

[0156] Through the above content, the energy transfer relationship and fault propagation potential between components can be accurately analyzed, and the propagation path of potential faults can be predicted in real time, which not only helps to identify the source of the fault in advance, but also can accurately predict the propagation speed and impact range of the fault, thereby providing a more scientific basis for decision-making. Through efficient energy flow distribution analysis and propagation path prediction, the downtime of the loader can be significantly reduced, maintenance costs can be reduced, and the reliability and safety of the equipment can be improved.

[0157] Fault assessment and graded warning in S4 include:

[0158] S41, Fault risk assessment: Based on the comprehensive data after multimodal data fusion and the results of energy flow fault propagation prediction, the fault risk is comprehensively assessed;

[0159] S42, fault warning signal generation: setting multi-level warning rules, comparing the results of the fault risk assessment with the preset risk threshold, and generating fault warning signals of different levels, including low risk warning, medium risk warning and high risk warning;

[0160] S43, generating maintenance suggestions: automatically generating corresponding maintenance suggestions according to the generated fault warning signals of different levels;

[0161] Through the above content, potential failure risks can be identified in real time and accurately, and targeted maintenance suggestions can be provided according to different risk levels. This can not only effectively prevent the occurrence of major failures and reduce equipment downtime, but also improve maintenance efficiency and reduce operating costs. At the same time, through the early warning mechanism, the allocation of maintenance resources is optimized, ensuring the efficient operation and safety of the loader, and improving the overall reliability and service life of the equipment.

[0162] Failure risk assessment in S41 includes:

[0163] S411, Fault risk assessment: Based on the comprehensive data after multi-modal data fusion and the results of energy flow fault propagation prediction, the fault risk index R is calculated. fault (i, t), expressed as:

[0164] R fault (i, t) = α1·S(i, t) + α2·P i,j (t)+α3·v i,j (t)+α4

[0165] ImpactRange(t);

[0166] Among them, R fault (i, t) represents the failure risk index of component i at time t, S(i, t) is the comprehensive data of component i at time t, and α1, α2, α3, and α4 are the corresponding weighting coefficients respectively;

[0167] S412, Comprehensively evaluate the failure risk of the loader: Use the failure risk index R of each component fault (i, t) is weighted and summed to evaluate the comprehensive failure risk index of the loader, which is expressed as:

[0168]

[0169] Among them, R total (t) is the comprehensive failure risk index of the loader at time t, ω i is the weight of component i;

[0170] Through the above content, not only the multi-faceted information of the equipment's operating status is taken into consideration, but also the dynamic process of fault propagation is combined, which can effectively identify the propagation trend and impact range of potential faults. Through multi-dimensional data fusion and comprehensive analysis of multiple factors, a more accurate risk prediction result is provided, which helps to identify potential fault hazards in advance, reduce unnecessary maintenance costs, optimize maintenance strategies, and improve the operating efficiency and reliability of loaders.

[0171] The fault warning signal generation in S42 includes:

[0172] S421, setting risk threshold: setting the risk threshold of loader failure risk, including the lower limit of the risk threshold R low and the upper risk threshold R high ;

[0173] Risk threshold lower limit R low and the upper risk threshold R high The settings are based on statistical analysis of historical data, including:

[0174] Collect historical data: Collect the operating status data of the loader in the past period of time, mark the failure events in the historical data, and calculate the historical comprehensive failure risk index for each failure event

[0175] Calculate the historical comprehensive failure risk index The statistical characteristics of are expressed as:

[0176]

[0177] Among them, μ R is the average value of the historical comprehensive failure risk index, σ R is the standard deviation of the historical comprehensive failure risk index, is the loader at time t i The historical comprehensive failure risk index at time, n is the number of time points;

[0178] Threshold setting: The lower limit of the risk threshold is set to the risk index value of most normal operation status in historical data, and the upper limit of the risk threshold is set to the risk index value close to the failure in historical data, expressed as:

[0179] R low =μ R -σ R ;

[0180] R high =μ R +σ R ;

[0181] S422, Risk level judgment and warning signal generation: Based on the fault risk assessment results and the risk threshold lower limit R low and the upper risk threshold R high By comparison, the fault risks are divided into different risk levels, and corresponding warning signals are generated according to different levels, including:

[0182] Low risk warning: When R total (t) <R low When , it indicates a low risk warning, and a low risk warning signal is generated;

[0183] Medium risk warning: When R low <R total (t) <R high When , it indicates a medium risk warning, and a medium risk warning signal is generated;

[0184] High risk warning: When R total (t)>R high When , it indicates a high-risk warning, and a high-risk warning signal is generated;

[0185] Through the above content, the risk level of potential failures can be accurately identified, and clear warning signals can be provided to maintenance personnel. Through refined risk assessment and warning signal classification, the accuracy and response speed of fault warnings can be effectively improved, thereby reducing the occurrence of equipment failures, optimizing the operation and maintenance management of loaders, and improving equipment reliability and work efficiency.

[0186] Maintenance suggestion generation in S43 includes:

[0187] S431, low-risk warning suggestion generation: based on the low-risk warning signal, low-risk warning suggestions are generated, including routine inspections, regular monitoring and data recording of components, and maintaining normal operation;

[0188] S432, medium risk warning suggestion generation: generating medium risk warning suggestions based on the medium risk warning signal, including inspecting the faulty parts (checking for abnormal signs or wear), arranging preventive maintenance of potential faulty parts, and inspecting the operating status of the loader and performing calibration or adjustment;

[0189] S433, high-risk warning suggestion generation: based on the high-risk warning signal, high-risk warning suggestions are generated, including immediate shutdown inspection and comprehensive fault diagnosis, and emergency repair or replacement of potential faulty parts;

[0190] Through the above content, the efficiency and accuracy of loader fault handling are improved, which not only improves the reliability and safety of the equipment, but also optimizes the allocation of maintenance resources and reduces operation and maintenance costs.

[0191] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0192] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A fault diagnosis and early warning method for loader multi-sensor data fusion, characterized in that: The following steps are involved: S1, multi-source data acquisition and preprocessing: using a variety of sensors arranged on the loader, real-time acquisition of the loader's operating status data, including vibration, temperature, pressure and current, and preprocessing of the acquired operating status data, including denoising, outlier correction, time alignment and standardization; S2, multimodal data fusion: the pre-processed operation status data is subjected to multimodal data fusion to form comprehensive data; S3, energy flow fault propagation prediction: Based on the power transmission path of the loader and the energy flow characteristics between the components, an energy flow fault propagation model is constructed to analyze the energy flow distribution of each component in real time and identify potential faulty components and their propagation paths, including: S31, constructing a loader power transmission path model: constructing an energy flow fault propagation model according to the loader power transmission path and energy flow characteristics between various components; S32, Energy flow distribution analysis: Based on the energy flow fault propagation model, the energy input, output and energy loss of each component are analyzed in real time, the operating efficiency and energy loss of each component are identified, and potential faults are discovered; S33, Fault propagation path prediction: Based on the results of energy flow distribution analysis, predict the propagation path strength, propagation rate and impact range of potential faults; S4, fault assessment and graded warning: Based on the integrated data after fusion and combined with the results of energy flow fault propagation prediction, the failure risk of the loader is comprehensively assessed, and graded warning signals are generated according to the set multi-level warning rules, including low-risk warning, medium-risk warning and high-risk warning, and corresponding maintenance suggestions are provided for different warning signals.

2. The fault diagnosis and early warning method of multi-sensor data fusion for loaders according to claim 1 is characterized in that: The multi-source data collection and preprocessing in S1 includes: S11, multi-source data collection: using a variety of sensors arranged on the loader, including vibration sensors, temperature sensors, pressure sensors and current sensors, to collect the operating status data of the loader in real time; S12, data preprocessing: preprocessing the real-time collected operation status data, specifically including: S121, denoising: using wavelet transform method to denoise the running status data, decomposing the original signal into multi-scale wavelet coefficients, and removing high-frequency noise signals; S122, outlier correction: detect and correct outliers by IQR method; S123, time alignment: In view of the sampling frequency differences of multiple sensors, the linear interpolation method is used to time align the operation status data and unify the timestamp; S124, normalization processing: normalizing the running status data.

3. The fault diagnosis and early warning method of multi-sensor data fusion for loaders according to claim 1 is characterized in that: The multimodal data fusion in S2 includes: S21, data mapping and normalization: mapping the preprocessed operating status data to a uniform value range through a normalization method; S22, weighted fusion of multimodal data: The normalized operating status data is subjected to a weighted fusion method to generate comprehensive data S(t).

4. The fault diagnosis and early warning method of multi-sensor data fusion for loaders according to claim 1 is characterized in that: The energy flow fault propagation model in S31 adopts a graph neural network model, and the graph neural network model includes: S311, constructing a graph structure model: modeling the power transmission path of the loader as a weighted directed graph G = (V, E), where V is a node set representing the components of the loader, E is an edge set representing the energy flow path between the components, and the weight represents the energy transmission efficiency, specifically including: Node and edge initialization: initial feature vector for each node It is the operation status data; Edge weight initialization: edge weight w i,j Initialized as energy transfer efficiency; S312, dynamic node feature update: update the feature vector of each node according to the node's neighbor information and edge weight, adding the time dimension and energy flow relationship; S313, fault propagation path calculation: Using the updated results of node features, calculate the fault propagation probability P of each edge i,j and time delay T i,j ; S314, global fault risk assessment: Based on the final characteristics of nodes and edges, the risk scope and critical path of global fault propagation are assessed, including: Global fault risk score: based on the fault propagation probability P of each edge i,j and time delay T i,j , calculate the global failure risk score R; Main propagation path: Select P i,j ·T i,j The largest path is used as the main propagation path for fault propagation.

5. The fault diagnosis and early warning method of multi-sensor data fusion for loaders according to claim 4 is characterized in that: The energy flow distribution analysis in S32 includes: S321, Calculation of energy input, output and energy loss: For each component i, the energy input E in,i , Energy output E out,i and energy loss E loss,i Perform calculations; S322, Evaluation of operating efficiency and energy loss: By comparing the input energy and output energy of each component, calculate its operating efficiency η i (t); S323, fault detection and early warning: perform fault detection based on the threshold of operating efficiency and energy loss. If η i (t) is lower than the set efficiency threshold or E loss,i If the loss threshold is exceeded, the component is considered to be potentially faulty.

6. The fault diagnosis and early warning method of multi-sensor data fusion for loaders according to claim 5 is characterized in that: The fault propagation path prediction in S33 includes: S331, fault propagation path strength calculation: by analyzing the energy input, output and loss of each component, combined with the energy flow fault propagation model, the energy transfer relationship and fault propagation potential between components are calculated, and the fault propagation path strength is predicted; S332, propagation rate calculation: based on the fault propagation path, analyze the propagation rate of the fault energy change per unit time from component i to component j; S333, Impact Scope Prediction: Predict the impact scope of the fault based on the propagation path and rate, and determine the impact scope of the fault by comprehensively evaluating the fault propagation potential of each component and the overall energy transfer efficiency.

7. The fault diagnosis and early warning method of multi-sensor data fusion for loaders according to claim 1, characterized in that: The fault assessment and graded warning in S4 include: S41, Fault risk assessment: Based on the comprehensive data after multimodal data fusion and the results of energy flow fault propagation prediction, the fault risk is comprehensively assessed; S42, fault warning signal generation: setting multi-level warning rules, comparing the results of the fault risk assessment with the preset risk threshold, and generating fault warning signals of different levels, including low risk warning, medium risk warning and high risk warning; S43, generating maintenance suggestions: automatically generating corresponding maintenance suggestions according to the generated fault warning signals of different levels.

8. The fault diagnosis and early warning method of multi-sensor data fusion for loaders according to claim 7 is characterized in that: The failure risk assessment in S41 includes: S411, Fault risk assessment: Based on the comprehensive data after multi-modal data fusion and the results of energy flow fault propagation prediction, the fault risk index R is calculated. fault (i, t); S412, Comprehensively evaluate the failure risk of the loader: Use the failure risk index R of each component fault (i, t) is weighted and summed to calculate the comprehensive failure risk index R total (t).

9. The fault diagnosis and early warning method of multi-sensor data fusion for loaders according to claim 8, characterized in that: The generation of the fault warning signal in S42 includes: S421, setting risk threshold: setting the risk threshold of loader failure risk, including the lower limit of the risk threshold R low and the upper risk threshold R high ; S422, Risk level judgment and warning signal generation: Based on the fault risk assessment results and the risk threshold lower limit R low and the upper risk threshold R high By comparison, the fault risks are divided into different risk levels, and corresponding warning signals are generated according to different levels, including: Low risk warning: When R total (t) <R low When , it indicates a low risk warning, and a low risk warning signal is generated; Medium risk warning: When R low <R total (t) <R high When , it indicates a medium risk warning, and a medium risk warning signal is generated; High risk warning: When R total (t)>R high When , it indicates a high-risk warning, a high-risk warning signal is generated.

10. The fault diagnosis and early warning method of multi-sensor data fusion for a loader according to claim 9, characterized in that: The maintenance suggestion generation in S43 includes: S431, low-risk warning suggestion generation: based on the low-risk warning signal, low-risk warning suggestions are generated, including routine inspections, regular monitoring and data recording of components, and maintaining normal operation; S432, medium risk warning suggestion generation: generating medium risk warning suggestions based on the medium risk warning signal, including checking the faulty parts, arranging preventive maintenance of potential faulty parts, checking the operating status of the loader and performing calibration or adjustment; S433, high-risk warning suggestion generation: Generate high-risk warning suggestions based on high-risk warning signals, including immediate shutdown inspection and comprehensive fault diagnosis, and emergency repair or replacement of potential faulty parts.

Images