Engineering machinery state monitoring method and device and engineering machinery

Abstract

The invention discloses an engineering machinery state monitoring method and device and engineering machinery. The accuracy and efficiency of engineering machinery state monitoring can be improved. The engineering machinery state monitoring method comprises the following steps: acquiring multi-dimensional sensor data acquired by a plurality of sensors; wherein one sensor corresponds to sensor data of one dimension; according to the sensor data of the multiple dimensions, determining the data correlation degree among the multiple sensors and the data correlation degree of the sensor data of the same dimension in different delay time; weighting the sensor data in the same dimension based on the data correlation degrees of the sensor data in the same dimension in different delay time to obtain weighted features; according to the data correlation degree among the plurality of sensors and the weighted features, constructing state monitoring features; wherein the change trend of the state monitoring characteristics reflects the operation state of the engineering machinery.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering machinery, and in particular to an engineering machinery state monitoring method and device, and engineering machinery. Background Art

[0002] In the field of engineering machinery, with the rapid development of science and technology, various large, complex and highly integrated mechanical equipment are widely used in infrastructure construction, mining, transportation and other industries. The performance stability and operating efficiency of engineering machinery are directly related to the progress, quality and safety of the project. However, traditional engineering machinery status monitoring methods often rely on manual inspections, regular maintenance and empirical judgment, which are inefficient. At present, the status of engineering machinery can be monitored based on models or based on statistical reliability prediction technology. However, both methods require a lot of time costs for preliminary training and statistics, and are not easy to iterate and optimize later, resulting in low efficiency. Summary of the invention

[0003] In order to solve the above technical problems, the present invention is proposed. The embodiments of the present invention provide a method and device for monitoring the state of an engineering machine, and an engineering machine, which can improve the accuracy and efficiency of monitoring the state of the engineering machine.

[0004] According to the first aspect of the present application, a method for monitoring the state of construction machinery is provided, comprising: acquiring sensor data of multiple dimensions collected by multiple sensors; wherein one sensor corresponds to sensor data of one dimension; determining, based on the sensor data of multiple dimensions, the degree of data correlation between the multiple sensors and the degree of data correlation of sensor data of the same dimension at different lag times; weighting the sensor data in the same dimension based on the degree of data correlation of the sensor data of the same dimension at different lag times to obtain weighted features; constructing a state monitoring feature based on the degree of data correlation between the multiple sensors and the weighted features; wherein the change trend of the state monitoring feature reflects the operating state of the construction machinery.

[0005] As a possible implementation manner, determining the degree of data correlation between the multiple sensors based on the sensor data of the multiple dimensions includes: determining a correlation matrix of the sensor data in the multiple sensors based on the sensor data of the multiple dimensions and grey correlation analysis; wherein the correlation matrix represents the degree of visual correlation of the multiple sensor data; based on the correlation matrix, clustering the multiple sensors; wherein each cluster includes a main sensor and at least one auxiliary sensor, and the auxiliary sensor is a sensor whose correlation degree with the main sensor reaches a preset value.

[0006] As a possible implementation method, based on the data correlation degree of the sensor data in the same dimension at different lag times, the sensor data in the same dimension is weighted to obtain a weighted feature, including: determining the autocorrelation coefficient based on the data correlation degree of the sensor data in the same dimension at different lag times; determining the maximum lag order of the autocorrelation coefficient of each sensor based on a preset confidence interval; wherein the maximum lag order represents the time period with the largest time interval in the autocorrelation coefficient; and determining the weighted feature of the maximum lag order based on the sensor data in the same dimension and the corresponding maximum lag order.

[0007] As a possible implementation method, the state monitoring feature is constructed according to the data correlation degree between the multiple sensors and the weighted feature, including: determining the main sensor parameters and the auxiliary sensor parameters in each cluster based on the main sensor and at least one auxiliary sensor in each cluster; constructing a multidimensional sensor sequence for each cluster based on the main sensor parameters and the auxiliary sensor parameters of each cluster; constructing a feature sequence based on the multidimensional sensor sequence of each cluster and the weighted features of the maximum lag order corresponding to the main sensor parameters; training a first self-organizing mapping network with the total number of the main sensor parameters and the auxiliary sensor parameters as the dimension of the feature sequence; taking the maximum value in each layer of the hidden matrix of the first self-organizing mapping network after training as the winning node value of the corresponding feature sequence; weightedly fusing the feature sequence and the winning node value corresponding to each time node to obtain the state monitoring feature.

[0008] As a possible implementation method, the method of determining the weighted feature of the maximum lag order based on the sensor data of the same dimension and the corresponding maximum lag order includes: training a second self-organizing map network based on the sensor data of the same dimension; inputting the sensor data of the same dimension into the trained second self-organizing map network to obtain network weights at different time points; the network weights represent weight vectors; determining the maximum weight value in the network weights as the winning node value at the corresponding time point; obtaining the winning node values ​​at all time points in a preset time series; using the maximum lag order as the time step and the winning node value in the time step as the weight, performing weighted calculation on the sensor data within the time step to obtain the weighted feature of the maximum lag order.

[0009] As a possible implementation method, the second self-organizing map network is trained based on sensor data of the same dimension, including: extracting the slope features of the sensor data in the same dimension to construct a two-dimensional time series; inputting the two-dimensional time series into the second self-organizing map network to train the network weights of the sensor; the second self-organizing map network clusters the different development trends of the two-dimensional time series according to the network weights; wherein the development trends include a gentle rise, a gentle fall, a sharp rise and a sharp fall; based on an internal metric for evaluating the clustering effect, the clustering effect is evaluated; when the evaluation score of the clustering effect is greater than or equal to a preset score, it is determined that the training of the second self-organizing map network is completed.

[0010] As a possible implementation method, after constructing the state monitoring feature according to the data correlation degree between the multiple sensors and the weighted feature, the construction machinery state monitoring method also includes: converting the state monitoring feature into a visual graph; wherein the visual graph includes a line graph; when a raised portion deviating from a preset fluctuation range appears in the line graph of the state monitoring feature, it is determined that a fault occurs in a time period corresponding to the raised portion deviating from the preset fluctuation range.

[0011] As a possible implementation method, after acquiring sensor data of multiple dimensions collected by multiple sensors, the construction machinery condition monitoring method also includes: performing frequency histogram analysis on the sensor data of the target dimension, and counting the frequency of occurrence of sensor data in different value domains; wherein the target dimension is any dimension among the multiple dimensions; based on the frequency of occurrence of sensor data in different value domains, determining the control upper limit and control lower limit of the value domain distribution of the sensor data in the target dimension; based on the control upper limit and control lower limit of the value domain distribution of the sensor data in the target dimension, normalizing the sensor data in the target dimension to obtain normalized target dimension data; determining the data correlation degree between the multiple sensors and the data correlation degree of sensor data of the same dimension at different lag times based on the sensor data of the multiple dimensions, includes: determining the data correlation degree between the multiple sensors and the data correlation degree of sensor data of the same dimension at different lag times based on the normalized sensor data of the multiple dimensions.

[0012] According to a second aspect of the present application, a construction machinery condition monitoring device is provided, comprising: an acquisition module, used to acquire sensor data of multiple dimensions collected by multiple sensors; wherein one sensor corresponds to sensor data of one dimension; a determination module, used to determine, based on the sensor data of multiple dimensions, the data correlation degree between the multiple sensors and the data correlation degree of sensor data of the same dimension at different lag times; a weighting module, used to weight the sensor data in the same dimension based on the data correlation degree of the sensor data of the same dimension at different lag times to obtain weighted features; a construction module, used to construct a condition monitoring feature based on the data correlation degree between the multiple sensors and the weighted feature; wherein the change trend of the condition monitoring feature reflects the operating status of the construction machinery.

[0013] According to the third aspect of the present application, there is provided a construction machinery, comprising: a plurality of sensors, the plurality of sensors being used to collect sensor data of multiple dimensions in the construction machinery; wherein one sensor collects sensor data of one dimension; and a monitoring device, the monitoring device being communicatively connected to the plurality of sensors, the monitoring device being used to execute the construction machinery status monitoring method described in the first aspect or any one of the implementation methods of the first aspect.

[0014] The engineering machinery state monitoring method, device and engineering machinery provided by the present invention can process multidimensional sensor data collected from complex system equipment, and construct state monitoring features based on the correlation between multidimensional sensors and the correlation between sensor data. The state monitoring features can intuitively reflect the state of the engineering machinery and improve the representativeness of the data. By comparing the state changes of the engineering machinery at different time points, the development trend of the state monitoring features over time can be obtained, so that the operating status of the engineering machinery can be judged by the fluctuation of the development trend, which effectively improves the efficiency and accuracy of engineering machinery state monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The above and other purposes, features and advantages of the present invention will become more apparent by describing the embodiments of the present invention in more detail in conjunction with the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 It is a flow chart of a method for monitoring the state of an engineering machinery provided by an exemplary embodiment of the present invention.

[0017] Figure 2 It is a schematic diagram of a flow chart of obtaining a maximum lag order weighted feature provided by an exemplary embodiment of the present invention.

[0018] Figure 3 is a schematic diagram of visualization of a construction status monitoring feature provided by an exemplary embodiment of the present invention.

[0019] Figure 4 It is a schematic structural diagram of an engineering machinery condition monitoring device provided by an exemplary embodiment of the present invention.

[0020] Figure 5 is a structural diagram of an electronic device provided by an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0021] Below, the exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described here.

[0022] The embodiments of the present invention can be applied to engineering machinery, which includes: multiple sensors for collecting sensor data of multiple dimensions in the engineering machinery; wherein one sensor collects sensor data of one dimension; and a monitoring device, which is communicatively connected to the multiple sensors, and the monitoring device is used to execute the engineering machinery status monitoring method provided by the present invention.

[0023] In some embodiments, the construction machinery may be an excavator, a crane, a transport machine, and the like.

[0024] In some embodiments, the sensor may involve sensors that collect engine speed, hydraulic oil temperature, cooling water temperature and pump pressure, wherein the collected engine speed can be regarded as sensor data of one dimension, and the hydraulic oil temperature can be regarded as sensor data of another dimension. The monitoring device obtains the sensor data collected by all sensors for constructing state monitoring features.

[0025] As an example, the multiple sensors may include an engine speed sensor (such as a crankshaft position sensor) for detecting the engine speed. The multiple sensors may include a hydraulic oil temperature sensor (such as a thermistor sensor or a thermocouple sensor) for detecting the temperature of the oil in the hydraulic system. The multiple sensors may include a cooling water temperature sensor (such as a thermistor sensor or an integrated temperature sensor) for detecting the engine cooling water temperature. The multiple sensors may include a pump pressure sensor (such as a capacitive sensor, a resistive sensor, a vibration sensor, or a static pressure sensor) for detecting the working pressure of the pump.

[0026] It should be understood that the above examples are only illustrative of sensors that collect engine speed, hydraulic oil temperature, cooling water temperature and pump pressure, but the embodiments of the present application are not limited to this. Other data similar to engine speed, hydraulic oil temperature, cooling water temperature and pump pressure can also be collected, such as fuel pressure, intake temperature, exhaust temperature, hydraulic oil pressure, vibration and noise levels, etc.

[0027] After collecting sensor data of various dimensions through multiple sensors, see Figure 1 The present invention can determine the data correlation between sensors of different dimensions and the data correlation between sensor data of the same dimension at different lag times based on sensor data of multiple dimensions (see Figure 1 In S100-S200 of FIG. 1 , based on the data correlation degree of the sensor data in the same dimension at different lag times, the sensor data in the same dimension are weighted to obtain weighted features (see Figure 1 Based on the data correlation degree and weighted features between multiple sensors, a condition monitoring feature is constructed (see S300 in Figure 1 The changing trend of the state monitoring feature is related to time. By observing the changing trend of the state monitoring feature, it is possible to quickly know whether there is an abnormal or prominent feature change in the engineering machinery. If the changing trend of the state monitoring feature continues to be stable without large fluctuations, it can be said that the state of the engineering machinery is stable and there is no abnormality. If there is an abnormal or prominent change in the changing trend of the state monitoring feature, it can be said that a fault occurs in the engineering machinery in the corresponding time period.

[0028] Combined with the following Figure 1 , a more detailed introduction is given to the engineering machinery status monitoring method provided in the embodiment of the present application.

[0029] In S100, sensor data of multiple dimensions collected by multiple sensors are obtained. Among them, one sensor corresponds to sensor data of one dimension. The sensors collect data of various dimensions on the engineering machinery. For example, the sensors may be engine speed sensors, hydraulic oil temperature sensors, cooling water temperature sensors, and pump pressure sensors, etc., and the corresponding sensor data collected may be engine speed, hydraulic oil temperature, cooling water temperature, and pump pressure, etc. In addition to the sensors and sensor data in the above examples, other sensors and sensor data related to engineering machinery may also be used.

[0030] In some implementations, one sensor corresponds to sensor data of one dimension. It can be understood that when the sensor is an engine speed sensor, the sensor data collected is the engine speed, which is sensor data of one dimension. When the sensor is a pump pressure sensor, the sensor data collected is the pump pressure, which is a sensor of another dimension. In other words, the data collected by one sensor is of one dimension, and the data collected by another sensor is of another dimension. Multiple sensors can collect sensor data of multiple dimensions.

[0031] Different types of sensors collect different sensor data, but these data are complementary in reflecting the state of construction machinery and can more comprehensively describe the current state of construction machinery. For example, multiple sensors may collect the same or similar data (such as temperature sensors at different locations), and these data can be mutually verified to improve the reliability and accuracy of the data. In addition, the state transition of construction machinery will trigger multiple sensors to collect sensor data at the same time. The sensor data collected by different sensors can be aligned on the time axis. Therefore, based on the time axis, multi-dimensional sensor data can be associated. By performing correlation analysis on the sensor data collected by multiple sensors, the inherent laws and connections of the state of construction machinery can be revealed, thereby judging the state of construction machinery. For example, by analyzing the correlation between data such as engine speed, oil pressure and oil temperature, it can be judged whether the working state of the engine is normal.

[0032] In S100 above, the raw sensor data collected by the sensor can be obtained, but different sensors usually have their own unique dimensions and ranges. For example, the data of a temperature sensor may be in Celsius, ranging from -50°C to 100°C; while the data of a pressure sensor may be in Pascal, ranging from 0 to 100000 Pa. Due to the differences in dimensions and ranges, directly comparing the data of different sensors may be meaningless or misleading.

[0033] In order to enable all sensor data to be integrated and compared with each other, in some embodiments, the sensor data of multiple dimensions collected by multiple sensors are normalized. That is, a frequency histogram analysis is performed on the sensor data of the target dimension, and the frequency of occurrence of sensor data in different value domains is counted; wherein the target dimension is any dimension among the multiple dimensions; based on the frequency of occurrence of sensor data in different value domains, the control upper limit and control lower limit of the value domain distribution of the sensor data in the target dimension are determined; based on the control upper limit and control lower limit of the value domain distribution of the sensor data in the target dimension, the sensor data in the target dimension is normalized to obtain normalized target dimension data. Correspondingly, S200 can be adjusted to: according to the normalized sensor data of multiple dimensions, determine the data correlation degree between multiple sensors and the data correlation degree of sensor data of the same dimension at different lag times.

[0034] Normalization involves scaling the data to a certain ratio so that it falls within a specific interval, usually the interval [0, 1], but it can also be other intervals, such as [-1, 1] or custom upper and lower bounds. The target dimension can represent the dimension collected by any sensor among multiple sensors. In the present invention, all sensor data involved in state monitoring need to be normalized, but the specific process of normalization is performed separately for each sensor.

[0035] As a normalization implementation method, for a single sensor, the implementation steps can be: 1. Perform frequency histogram analysis independently to count the frequency of sensor data (sensor data collected by a single sensor) in different value ranges; 2. Use the kernel density estimation method to estimate the probability density of each sensor data with the kernel as the weight and fit the distribution curve; 3. Use the trapezoidal numerical integration method to calculate the confidence interval of the kernel density curve (the confidence interval is obtained by subtracting two standard deviations from the sample mean), so as to determine the upper and lower control limits of the current sensor data value range distribution; 4. Normalize based on the upper and lower control limits of the sensor data. The sensor data collected by each sensor is normalized by the above steps 1 to 4, so that the sensor data of all sensors are normalized. Normalization can eliminate the influence of different sensor data due to different dimensions and value ranges, help accelerate the convergence speed of optimization algorithms such as gradient descent, reduce the imbalance of importance between different features, and improve the efficiency and accuracy of subsequent data application.

[0036] The so-called kernel density estimation (KDE) is a non-parametric statistical method used in probability theory to estimate unknown density functions. The kernel density estimation method does not rely on prior knowledge of data distribution and does not make any assumptions about data distribution. Instead, it studies the data distribution characteristics directly from the data sample itself, so it has strong applicability.

[0037] Continue to refer Figure 1 In S200, based on the sensor data of multiple dimensions, the data correlation degree between multiple sensors and the data correlation degree of the sensor data of the same dimension at different lag times are determined.

[0038] In some embodiments, different types of sensors can collect sensor data of different aspects of engineering machinery, such as temperature, pressure, vibration, and speed. These sensor data are complementary in reflecting the state of engineering machinery and can more comprehensively describe the operation of engineering machinery. Therefore, there is often a correlation between the sensor data collected by the sensors. For example, changes in engine speed may affect the temperature and pressure of the hydraulic oil, and the temperature and pressure of the hydraulic oil will affect the performance and stability of the engineering machinery. Sensor data of multiple dimensions can be combined in different ways and can be used to judge different states of engineering machinery. Therefore, determining the degree of data correlation between multiple sensors and selecting sensors with higher data correlation when monitoring the state of engineering machinery can improve the monitoring accuracy of the state of engineering machinery and more accurately judge the operating state and potential faults of engineering machinery.

[0039] As an example, when the sensors include an engine speed sensor, an accelerator position sensor, and a torque sensor, combined with the sensor data of the engine speed sensor, the accelerator position sensor, and the torque sensor, it is possible to evaluate whether the power output of the construction machinery is stable and whether it meets the current work requirements. When the sensors include a gearbox vibration sensor, an oil temperature sensor, and an oil pressure sensor, the operating status of the construction machinery transmission system, such as whether the gear meshing is good and whether the oil is sufficient, can be determined through the sensor data of the gearbox vibration sensor, the oil temperature sensor, and the oil pressure sensor. The vibration and noise levels of mechanical parts can be monitored by vibration sensors and sound sensors to determine their degree of wear. For example, the wear of bearings will lead to increased vibration and noise. Through data such as pressure sensors and displacement sensors, it can be determined whether the load condition of the construction machinery is appropriate.

[0040] A possible implementation of S200 may be: determining a correlation matrix of sensor data in multiple sensors based on sensor data of multiple dimensions and grey correlation analysis; wherein the correlation matrix represents the degree of visual correlation of multiple sensor data; based on the correlation matrix, clustering multiple sensors; wherein each cluster includes a main sensor and at least one auxiliary sensor, and the auxiliary sensor is a sensor whose correlation degree with the main sensor reaches a preset value.

[0041] The so-called Grey Relation Analysis (GRA) is a multi-factor statistical analysis method, which can be understood as, in a grey system, wanting to understand the relative strength of a certain indicator of interest being affected by other factors. Or, in a working condition, assuming that it is known that a certain indicator may be related to several other factors, and further wanting to know the relative strength of the correlation between this indicator and other factors, and so on, sorting these factors, and obtaining an analysis result, you can know which of the factors are more related to the indicator of interest. For example, analyze the similarity or difference between the trend development shapes of sensor data, and establish a correlation matrix based on grey correlation analysis. On the basis of the correlation matrix, each element and its relationship are visualized by a graph structure to reveal the spatial position relationship of each element. Each sensor saves the top N (for example, N = 4) other sensors with the highest correlation for clustering to optimize the visualized graph structure. According to the spatial position relationship visualized by the graph structure and the monitoring requirements, the main sensor and auxiliary sensor in each cluster are determined, and the number of auxiliary sensors is the same as N. The clustering method is not limited to grey relational analysis, but can also be hierarchical clustering, K-means clustering, density clustering, etc.

[0042] The so-called clustering of multiple sensors mainly refers to dividing physical or abstract objects into several groups (clusters) so that the objects in each group (cluster) have high similarity, that is, the sensor types within the same cluster have high correlation, while the similarity between objects in different groups (clusters) is relatively low. According to the distribution position of sensors between clusters, the main sensor type and auxiliary sensor type for constructing the state monitoring feature can be determined. For example, the speed, oil temperature and water temperature constitute a cluster. When constructing the speed state monitoring feature, the speed is selected as the main sensor parameter, and the oil temperature, water temperature, etc. are selected as auxiliary sensor parameters.

[0043] In order to intuitively view the clustering of sensors, as an example, the visualization of the graph structure can be a correlation heat map. In the first preset range (such as 0.80-1), the closer to 1, the darker the first color (such as dark blue), the darker the color, the stronger the correlation, and the farther away from 1, the lighter the first color (such as light blue), the lighter the first color, the lighter the correlation gradually weakens, and the associated values ​​in the second preset range are all set to red. In the second preset range (such as 0.60-0.75), the closer to 1, the lighter the second color (such as light red), and the farther away from 1, the darker the second color (such as dark red), the darker the second color, the darker the second color, the darker the correlation gradually weakens. Any value in the second preset range is less than the minimum value in the first preset range. Therefore, when selecting the main sensor and the auxiliary sensor, after determining the main sensor, the auxiliary sensor can be directly selected from the blue range.

[0044] In other embodiments, determining the data correlation degree of sensor data of the same dimension at different lag times refers to the correlation degree between the values ​​of a signal at different time points, and describes the strength and direction of the linear relationship between the current value and the past value or future value in the time series for the sensor data of the same dimension. By determining the data correlation degree of sensor data of the same dimension at different lag times, the stability of a certain dimension of the engineering machinery can be monitored. For example, for sensor data of a specified dimension, if the data correlation degree of the specified dimension remains within a relatively stable range, it means that the performance of the specified dimension of the engineering machinery at different time points is consistent and may be in a stable operating state. On the contrary, if the data correlation degree of the specified dimension fluctuates significantly, it may mean that there are unstable factors in the dimension in a certain period of time.

[0045] Therefore, after determining the data correlation between multiple sensors, more appropriate sensor data can be selected when judging the status, and the data correlation of sensor data of the same dimension at different lag times can be determined. When a fault occurs, the time point when the abnormality occurs can be found, and combined with other sensors with higher data correlation, the cause and time point of the fault can be jointly determined.

[0046] Continue to refer Figure 1 In S300, based on the data correlation degree of the sensor data in the same dimension at different lag times, the sensor data in the same dimension are weighted to obtain weighted features.

[0047] For a fault phenomenon or a state that needs to be monitored in a certain dimension, the importance of each single-dimensional sensor data is different. Therefore, by weighting the sensor data at different time points according to their importance, the data points that have a greater impact on the overall performance or prediction in the sensor data of the same dimension can be highlighted, thereby improving the representativeness of the data. Weighted features can help identify anomalies or mutation points in sensor data. Since the weights of anomalies are usually different from those of normal data points, they will also be more prominent in weighted features. In other words, when there is a fault phenomenon, the time point when the fault occurred can be more clearly determined through weighted features.

[0048] A possible implementation of S300 may be: determining the autocorrelation coefficient based on the degree of data correlation of sensor data of the same dimension at different lag times; determining the maximum lag order of the autocorrelation coefficient of each sensor based on a preset confidence interval; wherein the maximum lag order represents the time period with the largest time interval in the autocorrelation coefficient; determining the weighted feature of the maximum lag order based on the sensor data of the same dimension and the corresponding maximum lag order.

[0049] Autocorrelation, also called serial correlation, is the cross-correlation of a signal at different time points. Autocorrelation is a measure of how the similarity between observations of a sequence at different time points changes as the time interval (or lag) between them changes. In simple terms, it is a mathematical tool to evaluate the similarity between the current value and the past or future values ​​in the sequence, and this similarity depends on the time difference between them. The autocorrelation coefficient is a statistic that measures the degree of correlation between the values ​​of the same time series at different time points. The autocorrelation coefficient describes the strength and direction of the linear relationship between the current value and its past or future values ​​in the time series. The autocorrelation coefficient is used to analyze the data correlation of each sensor data at different lag times. This process is for single-dimensional sensors, and analyzes the data correlation of each sensor data at different lag times separately. For example, the range of the autocorrelation coefficient is [-1, 1]. When the autocorrelation coefficient is close to 1, it means that the time series is highly positively correlated with a lag of k periods. When the autocorrelation coefficient is close to -1, it means that the time series is highly negatively correlated with a lag of k periods. When the autocorrelation coefficient is close to 0, it means that the time series has almost no linear relationship with a lag of k periods. In autocorrelation, the correlation of sensor data decreases as the number of lags increases. The lag of autocorrelation refers to the number of time periods that separate the time series. For example, for a series with less than 240 observations, the maximum number of lags is about n / 4, and for a series with more than 240 observations, the maximum number of lags is Where n is the number of observations. For sequences with a large number of observations, the maximum lag number is usually calculated using something like The maximum lag order can be determined by empirical or software.

[0050] As an example of a possible implementation of S300, the maximum lag order when the autocorrelation coefficient of each sensor data is significantly not 0 can be determined based on the confidence interval. The maximum lag order refers to the one with the largest time interval (or lag) among all the autocorrelation coefficients that are significantly not 0. That is, determining the maximum lag order when the autocorrelation coefficient of each sensor data is significantly not 0 means that when a time point of the sensor data is compared with a certain time point in the past or future (this certain time point is the lag order), to what extent the similarity (i.e., the autocorrelation coefficient) between the observations at these two time points can be considered to be significantly not 0. Being significantly not 0 means that within a certain probability (determined by the confidence interval), it is believed that this autocorrelation coefficient is not caused by random errors, but is real.

[0051] The maximum lag order itself reflects the maximum time interval of significant correlation between observations in the data series. During this time interval, subsequent observations may be affected by previous observations. Therefore, the weighted features may adjust the original features according to this lag effect to better capture the dynamic changes and trends in the data. When constructing weighted features, weights can be assigned according to the autocorrelation coefficient corresponding to the maximum lag order. For example, if the autocorrelation coefficient of a lag order is high, it means that the similarity between the observations of the lag order and the current observation is high, so a higher weight can be given in the weighted feature. Conversely, if the autocorrelation coefficient is low or close to 0, a lower weight can be given or the influence of the lag order can be ignored.

[0052] A possible implementation method for determining the weighted feature of the maximum lag order may be: training a second self-organizing map network based on sensor data of the same dimension; inputting sensor data of the same dimension into the trained second self-organizing map network to obtain network weights at different time points; the network weights represent weight vectors; determining the maximum weight value in the network weights as the winning node value at the corresponding time point; obtaining the winning node values ​​at all time points in a preset time series; using the maximum lag order as the time step and the winning node value in the time step as the weight, performing weighted calculation on the sensor data within the time step to obtain the weighted feature of the maximum lag order.

[0053] The Self-Organizing Map (SOM) is a neural network for unsupervised learning. Its main feature is that it can map high-dimensional input data to a low-dimensional space (usually a two-dimensional or three-dimensional grid structure) while maintaining the original structure and relationship between the data. SOM can map complex high-dimensional data to a low-dimensional space, making data processing and analysis simpler and more efficient. By comparing the network weights at different time points, the dynamic changes of data in the low-dimensional space can be observed, thereby revealing the laws and trends hidden behind the data. Using the network weights generated by SOM, a distribution map or heat map of the data can be drawn to intuitively show the clustering of data at different time points, the distribution of outliers, etc., providing an intuitive visual reference for decision makers. In addition, by comparing the network weights at different time points, small changes in the data can be monitored. The clustering characteristics of SOM enable it to classify input data and identify outliers or outliers. By comparing the network weights at different time points, abnormal changes in the data can be more easily detected. The winning node value refers to the influence weight of the data that has the greatest impact on the fault or status at the current time point. Since determining the winning node value is a step that each unidimensional sensor needs to perform, the maximum lag order corresponding to each sensor is taken as the time step, and the winning node value at each time point in the time step is determined as the weight. The weighted calculation is performed to obtain the weighted feature of the maximum lag order of the unidimensional data. By weighting the data of different time steps according to their importance (i.e., the winning node value), the data points that have a greater impact on the overall performance or prediction can be highlighted, thereby improving the representativeness of the data. In time series analysis, weighted features can be used to predict future trends or behaviors. By focusing on past time steps that have an important impact on the future state of the system, the prediction of future trends can be improved.

[0054] As a more specific example of a possible implementation of S300, Figure 2 is a flow chart of obtaining the maximum lag order weighted feature provided by an exemplary embodiment of the present invention, see Figure 2 , divide the time series signal into multiple segments according to the maximum lag order k (divided into two segments in the figure), X1, X2, ... X k represents a time node, and each matrix W represents the weight of each sensor data at the corresponding time node. For example, the weight W1 of X1 is 0.962, 0.682, 0.037, and 0.998. The maximum value in each hidden layer matrix of the SOM network is taken as the winning node value of the corresponding feature sequence. Therefore, the winning node value of X1 is 0.998, and X k The weight W k are -0.419, 0.737, 0.906 and 0.938, where X kThe winning node value is 0.938. According to this method, the weight W of each time node in the first segment with the maximum lag order k is calculated. k , and then calculate the weighted features. Y1 represents the sensor data at time node X1, Y k Indicates time node X k Sensor data at time node X k The sensor data is multiplied by X k The weight of the winning node value is calculated, and the results of all time nodes are summed to obtain the weighted feature. The formula for calculating the weighted feature can be: Where Z1 represents the weighted feature with the maximum lag order k, the subscript i=1 indicates that the starting index of the summation is i=1, which is used to traverse each item in the sequence during the summation process. The superscript k indicates that the ending index of the summation is k, and the summation process will continue until i reaches k (including k), W i represents the i-th item in the sequence being summed. For each value of i (from 1 to k), the corresponding W i The value is added to the sum, Y k Indicates time node X k of sensor data.

[0055] As an example, based on sensor data of the same dimension, a method for training a second self-organizing map network may be: extracting the slope features of the sensor data in the same dimension to construct a two-dimensional time series; inputting the two-dimensional time series into the second self-organizing map network to train the network weights of the sensor; the second self-organizing map network clusters different development trends of the two-dimensional time series according to the network weights; wherein the development trends include a gentle rise, a gentle fall, a sharp rise, and a sharp fall; evaluating the clustering effect based on an internal metric for evaluating the clustering effect; and determining that the training of the second self-organizing map network is completed when the evaluation score of the clustering effect is greater than or equal to a preset score.

[0056] As a more specific example, for a single-dimensional sensor, the slope features are extracted independently to construct a two-dimensional time series. A time series refers to a sequence formed by arranging the values ​​of a statistical indicator at different times in a certain working condition in a timeline order. Initialize the SOM network with a hidden layer dimension of 2×2×2, input the two-dimensional time series into the SOM network, and independently train the network weights of sensors in each dimension.

[0057] As a more specific example, visualize the SOM training results and cluster the different development trends of the time series according to the SOM weights, including gentle rise, gentle fall, sharp rise, and sharp fall.

[0058] As a more specific example, the internal metric can be a CH score (Calinski-Harabasz index), and the clustering effect is evaluated by the CH score. The CH score evaluates the clustering result by calculating the ratio of the sum of squares of the distances between each point in the class and the center of the class (representing the closeness within the class) to the sum of squares of the distances between the center points of each class and the center points of the entire data set (representing the separation of the data set). The smaller the CH score, the poorer the clustering effect. If the clustering result is not good (less than the preset score), the network weights of the sensors in each dimension are returned to be retrained. The larger the CH score (greater than or equal to the preset score), the better the clustering effect, that is, the tighter the class itself, the more dispersed the classes are, and the second self-organizing map network training is determined to be completed. If the clustering result is good, the normalized single-dimensional data is input into the trained SOM network in sequence to obtain the network weight at the current time point. Find the maximum value in the network weight as the winning node value of the current network. Repeatedly input the normalized single-dimensional data into the trained SOM network, and find the maximum value in the obtained network weight until the winning node value of all time points in the current time series is obtained.

[0059] Continue to refer Figure 1 In S400, a state monitoring feature is constructed according to the data correlation degree and weighted features between the multiple sensors. The change trend of the state monitoring feature reflects the operation state of the construction machinery.

[0060] Weighted features refer to features that are weighted after calculating values ​​with prominent features in a single dimension. Constructing state monitoring features is actually a process of integrating data from multiple sources (multiple sensors) and time points to form more useful and comprehensive information. For example, through the data correlation between multiple sensors, it is found that when monitoring the speed state, oil temperature, water temperature, etc. have the greatest impact on the speed. When constructing the speed state monitoring feature, the speed, oil temperature and water temperature can be considered together. Then, the changes in speed, oil temperature and water temperature at different time points must also be considered. When constructing fusion features, the temporal correlation will also be considered, so that when an abnormal change trend occurs, the time point of the fault and the sensor data of the fault can be quickly located. When the change trend of the state monitoring feature is gentle, it can represent the normal operation of the current engineering machinery without any fault or abnormality. The change trend of the state monitoring feature shows a sharp rise and a sharp drop, which can indicate that there is a fault or abnormality in the engineering machinery. Therefore, the state monitoring feature extracted by the present invention can effectively and intuitively identify the abnormal state of the engineering machinery relative to the original sensor data, and the fused state monitoring feature can more comprehensively evaluate the abnormal state of the engineering machinery and improve the accuracy and efficiency of state monitoring.

[0061] A possible implementation of S400 may be: based on the main sensor and at least one auxiliary sensor in each cluster, determine the main sensor parameters and auxiliary sensor parameters in each cluster; construct a multidimensional sensor sequence for each cluster according to the main sensor parameters and auxiliary sensor parameters of each cluster; construct a feature sequence based on the multidimensional sensor sequence of each cluster and the weighted features of the maximum lag order corresponding to the main sensor parameters; train the first self-organizing map network with the total number of main sensor parameters and auxiliary sensor parameters as the dimension of the feature sequence; use the maximum value in each layer of the hidden matrix of the first self-organizing map network after training as the winning node value of the corresponding feature sequence; perform weighted fusion on the feature sequence and the winning node value corresponding to each time node to obtain the state monitoring feature.

[0062] As an example of a possible implementation of S400, according to the sensor correlation cluster determined above, the main sensor parameters and the auxiliary sensor parameters are used to construct a multidimensional sensor sequence. For example, to construct a speed state monitoring feature, the speed is the main sensor parameter, and the oil temperature, water temperature, etc. are auxiliary sensor parameters. The multidimensional sensor sequence is constructed into a feature sequence with the maximum lag order of the main parameter as a reference to obtain a multidimensional feature sequence, which can be understood as fusing multidimensional data based on a time series. Initialize a SOM network (first self-organizing map network) with a hidden layer dimension size of 15×15×K, where K is the dimension of the feature sequence, that is, the total number of main sensor parameters and auxiliary sensor parameters. Input the multidimensional feature sequence into the initialized SOM network for iterative training. Take the maximum value in each hidden layer matrix of the SOM network as the winning node value of the corresponding feature sequence. At each time node, the corresponding feature sequence is weightedly fused with the winning node value to construct a state monitoring feature.

[0063] In some embodiments, in order to facilitate intuitive viewing of the state of the engineering machinery and find faults therefrom, the state monitoring feature can be visualized. Therefore, after S400, the engineering machinery state monitoring method can also include: converting the state monitoring feature into a visualized graph; wherein the visualized graph includes a line graph; when a convex portion deviating from a preset fluctuation range appears in the line graph of the state monitoring feature, determining that a fault occurs in the time period corresponding to the convex portion deviating from the preset fluctuation range.

[0064] Taking the construction of speed status monitoring features as an example, Figure 3 is a schematic diagram of a visualization of a construction status monitoring feature provided by an exemplary embodiment of the present invention, see Figure 3The visualized line chart shows the main sensor parameter as speed, and the oil temperature and water temperature as auxiliary sensor parameters. The time period of 8834 in the figure is an abnormal time period. It can be seen that the status monitoring characteristics show an obvious abnormal trend in this time period. In addition, in other time periods except the 8834 time period, the speed characteristic alone shows an abnormal trend, but the water temperature characteristic and the oil temperature characteristic are stable in other positions, so the speed status monitoring characteristics do not show an abnormal trend as a whole. Between the 8900 time period and the 8950 time period, the water temperature characteristic showed an abnormal trend, but the speed characteristic and the oil temperature characteristic were relatively stable in this time period, so the speed status monitoring characteristic did not show an abnormal trend. Therefore, it can be seen that the speed status monitoring characteristic is abnormal in the position where the speed characteristic and the water temperature characteristic are abnormal, and the fault can be checked from the speed and water temperature respectively.

[0065] Figure 4 FIG. 1 is a schematic diagram of a structure of an engineering machinery state monitoring device provided by an exemplary embodiment of the present invention. Figure 4 As shown, the engineering machinery state monitoring device 4 includes: an acquisition module 41, which is used to acquire sensor data of multiple dimensions collected by multiple sensors; wherein one sensor corresponds to sensor data of one dimension; a determination module 42, which is used to determine the data correlation degree between multiple sensors and the data correlation degree of sensor data of the same dimension at different lag times based on the sensor data of multiple dimensions; a weighting module 43, which is used to weight the sensor data in the same dimension based on the data correlation degree of sensor data of the same dimension at different lag times to obtain weighted features; a construction module 44, which is used to construct a state monitoring feature according to the data correlation degree between multiple sensors and the weighted feature; wherein the change trend of the state monitoring feature reflects the operating state of the engineering machinery.

[0066] The engineering machinery state monitoring device provided by the present invention can process multidimensional sensor data collected from complex system equipment, and construct state monitoring features based on the correlation between multidimensional sensors and the correlation between sensor data. The state monitoring features can intuitively reflect the state of the engineering machinery and improve the representativeness of the data. By comparing the state changes of the engineering machinery at different time points, the development trend of the state monitoring features over time can be obtained, so that the operating status of the engineering machinery can be judged by the fluctuation of the development trend, which effectively improves the efficiency and accuracy of engineering machinery state monitoring.

[0067] In one embodiment, the determination module 42 can be configured to: determine a correlation matrix of sensor data in multiple sensors based on sensor data of multiple dimensions and grey correlation analysis; wherein the correlation matrix represents the degree of visual correlation of multiple sensor data; based on the correlation matrix, cluster the multiple sensors; wherein each cluster includes a main sensor and at least one auxiliary sensor, and the auxiliary sensor is a sensor whose correlation degree with the main sensor reaches a preset value.

[0068] In one embodiment, the weighting module 43 can be configured as follows: determining the autocorrelation coefficient based on the degree of data correlation of sensor data of the same dimension at different lag times; determining the maximum lag order of the autocorrelation coefficient of each sensor based on a preset confidence interval; wherein the maximum lag order represents the time period with the largest time interval in the autocorrelation coefficient; determining the weighted characteristics of the maximum lag order based on the sensor data of the same dimension and the corresponding maximum lag order.

[0069] In one embodiment, the construction module 44 can be configured as follows: based on the main sensor and at least one auxiliary sensor in each cluster, determine the main sensor parameters and auxiliary sensor parameters in each cluster; construct a multidimensional sensor sequence for each cluster according to the main sensor parameters and auxiliary sensor parameters of each cluster; construct a feature sequence based on the multidimensional sensor sequence of each cluster and the weighted features of the maximum lag order corresponding to the main sensor parameters; train the first self-organizing mapping network with the total number of main sensor parameters and auxiliary sensor parameters as the dimension of the feature sequence; use the maximum value in each layer of the hidden matrix of the first self-organizing mapping network after training as the winning node value of the corresponding feature sequence; and perform weighted fusion on the feature sequence and the winning node value corresponding to each time node to obtain the state monitoring feature.

[0070] In one embodiment, the weighting module 43 can also be configured as: training a second self-organizing mapping network based on sensor data of the same dimension; inputting sensor data of the same dimension into the trained second self-organizing mapping network to obtain network weights at different time points; the network weight represents a weight vector; determining the maximum weight value in the network weight as the winning node value at the corresponding time point; obtaining the winning node values ​​at all time points in a preset time series; using the maximum lag order as the time step and the winning node value in the time step as the weight, performing weighted calculation on the sensor data within the time step to obtain the weighted feature of the maximum lag order.

[0071] In one embodiment, the weighting module 43 can also be configured to: extract the slope characteristics of the sensor data in the same dimension to construct a two-dimensional time series; input the two-dimensional time series into the second self-organizing map network to train the network weights of the sensor; the second self-organizing map network clusters the different development trends of the two-dimensional time series according to the network weights; wherein the development trends include a gentle rise, a gentle fall, a sharp rise and a sharp fall; based on an internal metric for evaluating the clustering effect, the clustering effect is evaluated; when the evaluation score of the clustering effect is greater than or equal to a preset score, it is determined that the training of the second self-organizing map network is completed.

[0072] In one embodiment, the engineering machinery condition monitoring device 4 can also be configured to: convert the condition monitoring characteristics into a visual graph; wherein the visual graph includes a line graph; when a raised portion that deviates from a preset fluctuation range appears in the line graph of the condition monitoring characteristics, it is determined that a fault occurs in the time period corresponding to the raised portion that deviates from the preset fluctuation range.

[0073] In one embodiment, the engineering machinery condition monitoring device 4 can also be configured as follows: performing frequency histogram analysis on the sensor data of the target dimension, and counting the frequency of occurrence of sensor data in different value ranges; wherein the target dimension is any dimension among multiple dimensions; based on the frequency of occurrence of sensor data in different value ranges, determining the control upper limit and control lower limit of the value range distribution of the sensor data in the target dimension; based on the control upper limit and control lower limit of the value range distribution of the sensor data in the target dimension, normalizing the sensor data in the target dimension to obtain normalized target dimension data; wherein, based on the sensor data of multiple dimensions, determining the data correlation degree between multiple sensors and the data correlation degree of sensor data of the same dimension at different lag times, including: determining the data correlation degree between multiple sensors and the data correlation degree of sensor data of the same dimension at different lag times based on the normalized sensor data of multiple dimensions.

[0074] An electronic device comprises: a processor; a memory for storing instructions executable by the processor; and the processor is used to execute the engineering machinery state monitoring method described in the embodiment provided by the present invention.

[0075] Below, reference Figure 5 The electronic device according to the embodiment of the present invention is described. The electronic device may be any one or both of the first device and the second device, or a stand-alone device independent of them, and the stand-alone device may communicate with the first device and the second device to receive the collected input signals from them.

[0076] Figure 5 A block diagram of an electronic device according to an embodiment of the present invention is illustrated.

[0077] like Figure 5 As shown, the electronic device 10 includes one or more processors 11 and a memory 12 .

[0078] The processor 11 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.

[0079] The memory 12 may include one or more computer program products, and the computer program product may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 11 may run the program instructions to implement the engineering machinery condition monitoring method of each embodiment of the present invention described above and / or other desired functions. Various contents such as input signals, signal components, noise components, etc. may also be stored in the computer-readable storage medium.

[0080] In one example, the electronic device 10 may further include: an input device 13 and an output device 14, and these components are interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0081] When the electronic device is a stand-alone device, the input device 13 may be a communication network connector, which is used to receive the collected input signals from the first device and the second device.

[0082] In addition, the input device 13 may also include, for example, a keyboard, a mouse, and the like.

[0083] The output device 14 can output various information to the outside, including the determined distance information, direction information, etc. The output device 14 can include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto, and the like.

[0084] Of course, to simplify, Figure 5 Only some of the components related to the present invention in the electronic device 10 are shown, and components such as a bus, an input / output interface, etc. are omitted. In addition, according to specific application conditions, the electronic device 10 may also include any other appropriate components.

[0085] The computer program product may be written in any combination of one or more programming languages ​​to write program code for performing the operations of the embodiments of the present invention, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0086] A computer-readable storage medium stores a computer program, and the computer program is used to execute the engineering machinery condition monitoring method described in the embodiment provided by the present invention.

[0087] The computer readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0088] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present invention to the forms disclosed herein. Although a number of example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A method for monitoring the condition of construction machinery, characterized in that: include: Acquire sensor data of multiple dimensions collected by multiple sensors; wherein one sensor corresponds to sensor data of one dimension; Determining, based on the sensor data of the multiple dimensions, data correlation between the multiple sensors and data correlation between the sensor data of the same dimension at different lag times; Based on the data correlation degree of the sensor data in the same dimension at different lag times, weighting the sensor data in the same dimension to obtain a weighted feature; A state monitoring feature is constructed according to the data correlation degree between the multiple sensors and the weighted feature; wherein the change trend of the state monitoring feature reflects the operating state of the engineering machinery.

2. The method for monitoring the state of construction machinery according to claim 1, characterized in that: The step of determining the data correlation degree between the multiple sensors according to the sensor data of the multiple dimensions includes: Determine a correlation matrix of sensor data in the multiple sensors according to the sensor data of the multiple dimensions and the grey correlation analysis; wherein the correlation matrix represents the visual correlation degree of the multiple sensor data; Based on the association matrix, the multiple sensors are clustered; wherein each cluster includes a main sensor and at least one auxiliary sensor, and the auxiliary sensor is a sensor whose association degree with the main sensor reaches a preset value.

3. The method for monitoring the state of construction machinery according to claim 2, characterized in that: The step of weighting the sensor data in the same dimension based on the data correlation degree of the sensor data in the same dimension at different lag times to obtain the weighted feature includes: Determining an autocorrelation coefficient based on the data correlation degree of the sensor data of the same dimension at different lag times; Based on a preset confidence interval, determining a maximum lag order of the autocorrelation coefficient of each sensor; wherein the maximum lag order represents a time period with the largest time interval in the autocorrelation coefficient; Based on the sensor data of the same dimension and the corresponding maximum lag order, a weighted feature of the maximum lag order is determined.

4. The method for monitoring the state of construction machinery according to claim 3, characterized in that: The constructing of the state monitoring feature according to the data correlation degree between the multiple sensors and the weighted feature includes: Determining a primary sensor parameter and an auxiliary sensor parameter in each cluster based on the primary sensor and at least one auxiliary sensor in each cluster; constructing a multidimensional sensor sequence for each cluster according to the main sensor parameters and the auxiliary sensor parameters of each cluster; Constructing a feature sequence based on the multidimensional sensor sequence of each cluster and the weighted features of the maximum lag order corresponding to the main sensor parameters; Training a first self-organizing map network using the total number of the main sensor parameters and the auxiliary sensor parameters as the dimension of the feature sequence; Taking the maximum value in each layer of the hidden matrix of the first self-organizing map network after training as the winning node value of the corresponding feature sequence; The feature sequence corresponding to each time node and the winning node value are weightedly fused to obtain the state monitoring feature.

5. The method for monitoring the state of construction machinery according to claim 3, characterized in that: The step of determining the weighted feature of the maximum lag order based on the sensor data of the same dimension and the corresponding maximum lag order includes: Based on the sensor data of the same dimension, training a second self-organizing map network; Inputting sensor data of the same dimension into the trained second self-organizing map network to obtain network weights at different time points; the network weights represent weight vectors; Determine the maximum weight value in the network weight as the winning node value at the corresponding time point; Get the winning node value at all time points in the preset time series; The maximum lag order is used as the time step, and the winning node value in the time step is used as the weight, and weighted calculation is performed on the sensor data within the time step to obtain the weighted feature of the maximum lag order.

6. The method for monitoring the state of construction machinery according to claim 5, characterized in that: The training of the second self-organizing map network based on the sensor data of the same dimension includes: Extract the slope features of sensor data in the same dimension to construct a two-dimensional time series; Inputting the two-dimensional time series into a second self-organizing map network to train the network weights of the sensor; The second self-organizing map network clusters different development trends of the two-dimensional time series according to the network weights; wherein the development trends include a gentle rise, a gentle fall, a sharp rise and a sharp fall; Evaluate the clustering effect based on internal metrics for evaluating clustering effect; When the evaluation score of the clustering effect is greater than or equal to a preset score, it is determined that the second self-organizing map network training is completed.

7. The method for monitoring the state of construction machinery according to claim 1, characterized in that: After constructing the condition monitoring feature according to the data correlation degree between the multiple sensors and the weighted feature, the construction machinery condition monitoring method further includes: Converting the condition monitoring feature into a visual graph; wherein the visual graph comprises a line graph; When a convex portion deviating from the preset fluctuation range appears in the line graph of the state monitoring feature, it is determined that a fault occurs in the time period corresponding to the convex portion deviating from the preset fluctuation range.

8. The method for monitoring the state of construction machinery according to claim 1, characterized in that: After acquiring sensor data of multiple dimensions collected by multiple sensors, the construction machinery condition monitoring method further includes: Performing frequency histogram analysis on the sensor data of the target dimension to count the frequency of occurrence of the sensor data in different value domains; wherein the target dimension is any dimension among the multiple dimensions; Based on the frequency of occurrence of sensor data in different value ranges, determine the upper and lower control limits of the value range distribution of sensor data in the target dimension; Based on the upper control limit and the lower control limit of the value range distribution of the sensor data in the target dimension, the sensor data in the target dimension is normalized to obtain normalized target dimension data; The step of determining, based on the sensor data of the multiple dimensions, the data correlation degree between the multiple sensors and the data correlation degree of the sensor data of the same dimension at different lag times includes: According to the normalized sensor data of multiple dimensions, the data correlation degree between the multiple sensors and the data correlation degree of the sensor data of the same dimension at different lag times are determined.

9. A construction machinery condition monitoring device, characterized in that: include: An acquisition module, used to acquire sensor data of multiple dimensions collected by multiple sensors; wherein one sensor corresponds to sensor data of one dimension; A determination module, configured to determine, based on the sensor data of the multiple dimensions, the data correlation degree between the multiple sensors and the data correlation degree of the sensor data of the same dimension at different lag times; A weighting module, configured to weight the sensor data in the same dimension based on the data correlation degree of the sensor data in the same dimension at different lag times to obtain a weighted feature; A construction module is used to construct a state monitoring feature according to the data correlation degree between the multiple sensors and the weighted feature; wherein the change trend of the state monitoring feature reflects the operating state of the engineering machinery.

10. An engineering machine, characterized in that: include: A plurality of sensors, wherein the plurality of sensors are used to collect sensor data of multiple dimensions in the engineering machinery; wherein one sensor collects sensor data of one dimension; A monitoring device, wherein the monitoring device is communicatively connected to the plurality of sensors, and the monitoring device is used to execute the engineering machinery condition monitoring method described in any one of claims 1 to 8.

Images