Method and device for detecting abnormal running state of loading arm

Abstract

The invention discloses a crane tube operation state abnormity detection method and device, and the method comprises the steps: collecting crane tube operation parameters in real time in the crane tube operation process; fusing the operation parameters to obtain fused data; inputting the fusion data into a pre-trained anomaly detection model, and obtaining a reconstruction error according to the output of the anomaly detection model; and determining whether the crane tube operation state is abnormal according to the reconstruction error. By means of the scheme, the running state of the crane pipe can be detected in real time, and the accuracy and reliability of anomaly detection are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial automation and equipment monitoring, and in particular to a method and a device for detecting abnormal operation status of a crane pipe. Background Art

[0002] Crane pipe is a key loading and unloading equipment widely used in the storage and transportation industry, mainly used for loading and unloading of hazardous chemicals such as liquefied petroleum gas (LPG) and liquefied natural gas (LNG). Due to its complex operating environment and flammable and explosive transmission media, any abnormal operation or leakage failure of the equipment may cause serious safety accidents and threaten personal and property safety.

[0003] At present, traditional crane pipe anomaly detection methods usually rely on threshold judgment or rule setting. This method has poor adaptability to complex working conditions and is difficult to effectively detect early leakage or small-scale anomalies. At the same time, facing high-dimensional heterogeneous data collected by multiple sensors, traditional methods are difficult to fully explore the implicit correlation features in the data, which can easily lead to missed reports or false reports. In addition, noise interference in operating data may also reduce the reliability and accuracy of detection. Summary of the invention

[0004] The present invention provides a method and device for detecting abnormal operation status of a crane pipe, so as to effectively process multi-dimensional and highly complex data, realize real-time detection of the operation status of the crane pipe, and improve the accuracy and reliability of abnormal operation status detection of the crane pipe.

[0005] To this end, the present invention provides the following technical solutions: The present invention provides a method for detecting abnormal operation status of a crane pipe, the method comprising: During the operation of the crane pipe, the crane pipe operation parameters are collected in real time; fusing the operating parameters to obtain fused data; Inputting the fused data into a pre-trained anomaly detection model, and obtaining reconstructed data according to an output of the anomaly detection model; Calculating a reconstruction error based on the reconstructed data and the corresponding fused data; Whether the crane operation state is abnormal is determined according to the reconstruction error.

[0006] Optionally, the process of training an autoencoder-based anomaly detection model includes: Collect the working parameter sequence of the crane pipe based on the time series; Cleaning the working parameter sequence to obtain a cleaned data sequence; Normalizing each operating parameter in the data sequence; fusing a plurality of heterogeneous data at each time point in the data sequence to obtain a fused data sequence; Design the network structure of the autoencoder-based anomaly detection model and determine the loss function; The fused data sequence is used as training data, and model parameters are trained based on the loss function.

[0007] Optionally, the operating parameters include: temperature, flow, and pressure data.

[0008] Optionally, the cleaning of the working parameter sequence includes any one or more of the following: Processing abnormal values ​​and missing values ​​in the working parameter sequence; A noise reduction process is performed on the working parameter sequence.

[0009] Optionally, fusing the plurality of heterogeneous data at each time point in the data sequence includes: Determining the weight of each operating parameter in the data sequence; A weighted fusion value of multiple heterogeneous data at each time point in the data sequence is calculated according to the weight of each working parameter.

[0010] Optionally, determining the weight of each operating parameter in the data sequence includes: The weight of each working parameter in the data sequence is determined based on the following principles: (1) Dynamically adjust weights based on data reliability; (2) Assign weights based on the sensitivity of sensors to changes in variables.

[0011] Optionally, the autoencoder includes: an encoder and a decoder; the encoder includes a series of fully connected layers or convolutional layers that are reduced layer by layer; and the decoder includes a layer-by-layer expansion network that is symmetrical to the encoder.

[0012] Optionally, the using the fused data sequence as training data and training model parameters based on the loss function further comprises: Dividing the fused data sequence into a training set and a test set; Training model parameters using the training set and the loss function; The test set is used to evaluate the model effect, and the network structure and parameters of the model are optimized based on the evaluation results.

[0013] Optionally, determining whether the crane operation state is abnormal according to the reconstruction error includes: If the reconstruction error is greater than a set threshold, it is determined that the crane operation state is abnormal; Otherwise, make sure the crane is operating normally.

[0014] Optionally, the method further comprises: According to the reconstruction error of normal operation data, the error distribution is statistically analyzed; The set threshold is determined according to the error distribution.

[0015] Optionally, the method further includes: when an abnormal operation status of the crane pipe is detected, triggering an alarm, and recording the abnormal information in a log file or a database.

[0016] The present invention also provides a device for detecting abnormal operation status of a crane pipe, the device comprising: Data acquisition module, used to collect crane operation parameters in real time during the operation of the crane; A data fusion module, used for fusing the operating parameters to obtain fused data; A reconstruction error determination module, used for inputting the fused data into a pre-trained autoencoder-based anomaly detection model, and obtaining a reconstruction error according to an output of the anomaly detection model; The judgment module is used to determine whether the crane operation state is abnormal according to the reconstruction error.

[0017] Optionally, the device further comprises: a model building module, for training an anomaly detection model based on an autoencoder; The model building module includes: A data acquisition unit, used for acquiring a time series-based working parameter sequence of the crane pipe; A preprocessing unit, used for cleaning the working parameter sequence to obtain a cleaned data sequence; A normalization processing unit, used for performing normalization processing on each working parameter in the data sequence; A fusion unit, used for fusing a plurality of heterogeneous data at each time point in the data sequence to obtain a fused data sequence; The network structure design unit is used to design the network structure of the autoencoder-based anomaly detection model and determine the loss function; A training unit is used to use the fused data sequence as training data and train model parameters based on the loss function.

[0018] Optionally, the device further includes: an exception handling module, which is used to trigger an alarm when an abnormal operation status of the crane pipe is detected, and to record the abnormal information in a log file or a database.

[0019] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the crane pipe operation status abnormality detection method are executed.

[0020] The method and device for detecting abnormal operation status of crane pipes provided by the present invention utilize the unsupervised learning capability of the autoencoder model to model the normal operation mode of the crane pipes and identify abnormal situations by reconstructing errors. Compared with traditional methods, the scheme of the present invention can effectively process multi-dimensional and highly complex data, detect abnormal operation status of equipment in real time, and combine with early warning mechanisms to significantly improve the accuracy and reliability of abnormal operation status detection of crane pipes, thereby providing technical guarantee for the safe loading and unloading of hazardous chemicals.

[0021] Furthermore, by dynamically adjusting the fusion weights and optimizing the network structure, the detection accuracy and adaptability of the model are improved, making it suitable for industrial process monitoring and fault prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 is a flowchart of constructing an anomaly detection model in an embodiment of the present invention; Figure 2 is a structural diagram of an anomaly detection model based on an autoencoder in an embodiment of the present invention; Figure 3 This is a flow chart of a method for detecting abnormal operation status of a crane pipe according to an embodiment of the present invention; Figure 4 This is a structural schematic diagram of a device for detecting abnormal operation status of a crane pipe according to an embodiment of the present invention; Figure 5 It is a structural schematic diagram of a model building module in an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.

[0025] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0026] In view of the problems that the existing technology of abnormal operation status detection of crane pipes is not real-time enough, has low accuracy and poor adaptability to complex working conditions, the embodiments of the present invention provide a method and device for detecting abnormal operation status of crane pipes, which can efficiently and reliably detect data anomalies during the operation of crane pipes, especially dangerous conditions such as liquefied gas leakage, provide a real-time early warning mechanism for equipment operation, and ensure safe production.

[0027] In view of the problem that the traditional crane pipe anomaly detection method is difficult to fully explore the implicit correlation features in the data, which easily leads to missed reports or false reports, the embodiment of the present invention provides a crane pipe operation status anomaly detection method and device, which uses the unsupervised learning ability of the autoencoder model to model the normal mode of crane pipe operation and identify abnormal situations through reconstruction errors. Specifically, the autoencoder-based anomaly detection model is established using the time series-based working parameter sequence under the normal operation state of the collected crane pipe, and the crane pipe operation status is detected using the anomaly detection model.

[0028] First, the construction process of the anomaly detection model in the embodiment of the present invention is described in detail below.

[0029] like Figure 1 FIG. 1 is a flowchart of constructing an anomaly detection model in an embodiment of the present invention, which includes the following steps: In step 101, a time-series-based sequence of working parameters of the crane is collected.

[0030] The working parameters may include, but are not limited to, temperature, flow rate, and pressure data, etc., which are used as research parameters for abnormal detection of crane pipes.

[0031] Taking the crane pipe used by chemical companies as an example, sensors are placed at key nodes of the liquefied petroleum gas pipeline, such as the inlet, outlet and valve area in the middle. DN15-150 electromagnetic flowmeter, pressure transmitter and temperature transmitter are used to collect sensor data at a sampling frequency of 1 second / time. The data recording time range is from September 1, 2024 to September 15, 2024, and a total of about 1 million temperature, flow and pressure data are collected.

[0032] These operating parameters can be collected by sensors installed on the crane pipe.

[0033] In step 102, the working parameter sequence is cleaned to obtain a cleaned data sequence.

[0034] Cleaning the working parameter sequence mainly involves detecting and processing abnormal values ​​and missing values ​​therein to improve data quality and reliability. For example, in some embodiments, data cleaning may include any one or more of the following: (1) Processing abnormal values ​​in the working parameter sequence.

[0035] Specifically, any of the following methods can be used to detect outliers: Statistical method: Calculate the mean and standard deviation of the collected data, and mark the values ​​that exceed a certain range (for example, 3 times) of the standard deviation as abnormal.

[0036] Box plot method: Calculate the upper and lower limits based on the interquartile range (IQR): ; ; in, and The 25th and 75th percentiles, respectively. is the interquartile range.

[0037] The detected outliers can be processed by correction, elimination, marking and other methods, among which: Correction can be replacing outliers with the mean or median; Elimination can be to delete the data points corresponding to outliers; Marking can be to record outliers separately for subsequent analysis.

[0038] For example, using statistical methods to detect outliers: the mean of all temperature data is calculated to be 45°C, the standard deviation is 5°C, and it is found that there are many temperature data exceeding 45°C on September 7, 2024. The values ​​with a high sensitivity (<30°C or >60°C) were marked as outliers.

[0039] For another example, the box plot method is used to detect outliers in the flow data. It is found that 10 records in the flow data exceed the upper and lower limits and are determined to be outliers.

[0040] Then, all detected outliers are processed using mean filling (pressure data) and removal (temperature and flow data) strategies, respectively.

[0041] (2) Processing the missing values ​​in the working parameter sequence.

[0042] Specifically, different processing methods can be used according to different types of missing values. First, the type of missing values ​​is determined, for example, the number and distribution of missing values ​​in the collected data can be counted; if the proportion of missing values ​​is low, interpolation methods are used to fill them, including linear interpolation, spline interpolation, etc.; if the missing values ​​are concentrated in certain sensors, the sensor data can be removed or the mean filling method can be used; if there is a whole column missing in the data table (the data table is a data table generated based on data collected by different sensors), the column data is removed.

[0043] Among them, the interpolation algorithm can use the following formula to achieve linear interpolation:

[0044] in, and are the values ​​of the two most recent valid data points before and after the missing value, and is the corresponding timestamp.

[0045] For example, it is found that the pressure sensor has a high proportion of missing values ​​during the period of 12:00-12:30 on September 3, 2024, accounting for 20%. Then the missing pressure values ​​in this time period are filled by the above interpolation calculation. If the missing value ratio is higher than 30% or the entire column of data is missing, for example, the temperature sensor is missing the entire column on September 3, then the data in this time period is removed and the specific reason is recorded.

[0046] (3) Performing noise reduction processing on the working parameter sequence.

[0047] The specific noise reduction method may adopt some noise reduction methods in the prior art, for example, any of the following noise reduction methods may be adopted: Median filtering: Taking the median instead of the center point in the sliding window to remove outliers can effectively remove mutation noise and retain the edge characteristics of the signal.

[0048] Gaussian filtering: Use the weights defined by the Gaussian distribution function to perform weighted averaging on the data, which can preserve the overall trend of the signal.

[0049] Use median filtering to remove sudden noise, and then further smooth the signal through Gaussian filtering or wavelet transform.

[0050] Furthermore, the original un-denoised data can be retained for comparative analysis while the denoising process is being performed. If anomalies cannot be detected after denoising, the original data can be traced back for verification. When training the autoencoder, the distribution of reconstruction errors of the data before and after denoising is monitored to ensure that denoising does not significantly reduce sensitivity to anomalies.

[0051] In step 103, each operating parameter in the data sequence is normalized.

[0052] For example, the MinMaxScaler method can be used to normalize the data to the interval [0,1]. The calculation formula of MinMaxScaler is:

[0053] in, represents the normalized data value, Represents the original data value, Indicates the minimum value of the data. Indicates the maximum value of the data.

[0054] It should be noted that when performing normalization processing, it is necessary to perform normalization for different types of working parameters. For example, for temperature parameters, the maximum and minimum temperature values ​​in the data sequence are obtained, and then the corresponding normalized value of each temperature parameter is calculated using the above formula.

[0055] In step 104, a plurality of heterogeneous data at each time point in the data sequence are fused to obtain a fused data sequence.

[0056] The heterogeneous data refers to data of different types, that is, different working parameters.

[0057] In some embodiments, a weighted calculation may be performed on multiple heterogeneous data at each time point in the data sequence to obtain a weighted fusion value.

[0058] First, the weight of each operating parameter in the data sequence needs to be determined.

[0059] For example, the weight of each operating parameter in the data sequence may be determined based on the following principles: (1) Dynamically adjust weights based on data reliability: assign lower weights to sensor data with greater noise, and dynamically update the assigned weights based on indicators such as signal-to-noise ratio or data consistency; (2) Assign weights based on the sensitivity of sensors to changes in variables.

[0060] In order to reflect the response degree of the sensor to the change of the target variable, a sensitivity index can be defined :

[0061] in, For sensor The collected value, is the target variable, For sensor For the target variable sensitivity to changes.

[0062] Normalize the sensitivity indexes, that is, normalize all sensitivity indexes to the range of [0, 1]:

[0063] in, represents the normalized sensitivity index value, represents the original sensitivity index, represents the minimum value of the sensitivity index, Indicates the maximum value of the sensitivity index.

[0064] Assign weights based on sensitivity: The higher the sensitivity of the sensor, the faster the response to changes in the target variable and the greater the weight, which can be calculated using the following formula:

[0065] in, Represents the normalized The sensitivity index value of each sensor; It is the sum of the normalized sensitivity index values ​​of all sensors, which is used to normalize the weights to ensure that the sum of the weights is 1.

[0066] Then, the weighted fusion value of the plurality of heterogeneous data at each time point in the data sequence is calculated according to the weight of each working parameter. , weighted calculation is performed on the data of each sensor according to the weight, and the weighted fusion value of multiple heterogeneous data at that time point is obtained : ;or, ; in, For the Sensors at time The collected values ​​(i.e. the collected working parameter values), For the corresponding The weight of each sensor (i.e. the weight of the working parameter), is the target value after fusion.

[0067] In some embodiments, the fusion results can also be stored in a feature database (time series database) in the form of a time series to facilitate the subsequent use of these data, such as anomaly detection, trend analysis, prediction model establishment, etc. The feature data in the feature database (i.e., the weighted fusion value) is stored in the form of a time series.

[0068] In step 105, the network structure of the autoencoder-based anomaly detection model is designed and the loss function is determined.

[0069] like Figure 2 As shown, the autoencoder includes: an encoder and a decoder; the input of the anomaly detection model is the fused data in the form of a time series, that is, the fused data sequence obtained in the above step 104; the output of the anomaly detection model is the reconstruction error of the input data.

[0070] Figure 2In the example, the encoder input is X, and the encoder's goal is to convert it into a low-dimensional feature representation h. The decoder reconstructs an output r similar to the input based on the low-dimensional feature h output by the encoder, and loss is the loss value.

[0071] The encoder and decoder parts are described in detail below.

[0072] Encoder part (Encoder): It includes a series of fully connected layers or convolutional layers that are gradually reduced to extract the potential features of the data. The number of nodes in each layer decreases, such as [64, 32, 16], that is, the number of nodes in the input layer is 64 (feature dimension), the number of nodes in the first hidden layer is 32, and the number of nodes in the second hidden layer is 16, gradually compressing the data into a low-dimensional feature space. The encoder can use activation functions, such as ReLU (Rectified Linear Unit) or Leaky ReLU, to enhance the nonlinear expression ability of the network.

[0073] The encoder maps the input data to a low-dimensional latent space, removes redundant information, and retains only important features. Assuming the input is X, the goal of the encoder is to convert it into a low-dimensional feature representation h:

[0074] in, is the encoder mapping function, usually composed of several fully connected neural network layers.

[0075] Decoder part (Decoder): It includes a layer-by-layer expansion network that is symmetrical to the encoder to reconstruct the original data from the latent features. Ensure that the output dimension of the decoder is consistent with the input data. The decoder is based on the low-dimensional features output by the encoder. Reconstruct output similar to input :

[0076] In some embodiments, the mean square error (MSE) may be selected but not limited to as a loss function to measure the difference between the reconstructed data and the original input, and the formula is as follows:

[0077] in, is the original input data, Reconstructed data for the model.

[0078] For scenarios with more noise, we can further introduce weighted loss functions or robust loss functions (such as HuberLoss) to reduce the impact of abnormal data points on training.

[0079] Continue to refer to Figure 1 , in step 106, the fused data sequence is used as training data, and model parameters are trained based on the loss function.

[0080] The corresponding time series data can be extracted from the feature database as model input. For example, if the sampling frequency is 1 second / time, the data input format is ,in is the number of samples, is the time step, is the feature dimension.

[0081] The training of the autoencoder-based anomaly detection model can use some existing optimization algorithms, such as Adam (Adaptive Moment Estimation), RMSProp (Root Mean Square Propagation), or SGD (Stochastic Gradient Descent) to accelerate convergence. Adjust the learning rate (such as 0.001 or 0.0001) to balance the convergence speed and stability.

[0082] The training batch and training cycle can be set as needed. Setting an appropriate batch size (such as 32 or 64) ensures that the model can effectively learn features during training while improving computational efficiency. Determining the number of training iterations (Epoch) combined with the validation set early stopping mechanism (Early Stopping) can effectively avoid overfitting or underfitting of the model.

[0083] For example, in some embodiments, the Adam optimization algorithm is selected, and the initial learning rate is set to 0.001. A dynamic learning rate adjustment strategy is adopted to reduce the learning rate (e.g., halve it) when the performance of the validation set does not improve. The batch size is set to 32, and the training data is input into the model in batches during the experiment. The training cycle (Epoch) is set to 50, and combined with the early stopping mechanism (EarlyStopping), the training is stopped when the validation set loss does not decrease for 10 consecutive Epochs.

[0084] To further ensure the quality of the model, you can also verify and tune the trained anomaly detection model. For example, you can verify and tune it in the following ways: (1) Divide the data set into training set, validation set and test set to ensure that the training and validation data of the model are independent and make the model more robust. The validation set is used to evaluate the model performance, and the test set is used to test the model generalization ability.

[0085] By training the autoencoder model on normal operating data, it learns to reconstruct the feature distribution of the normal state. Since the model does not directly learn abnormal data, the reconstruction error is low for normal data but high for abnormal data. During the training process, the validation set is used to monitor the model performance to ensure that the reconstruction error remains stable on normal data and avoid overfitting or underfitting of the model.

[0086] (2) Evaluate the reconstruction performance of the model: Use the validation set to calculate the reconstruction error, evaluate the learning effect of the model on the normal operation data, adjust the number of network layers, nodes and activation functions, and improve the model performance. For example, test the impact of different compression rates ([64, 32, 16], [64, 16]) on model performance.

[0087] (3) Tuning strategy: Optimize hyperparameters (such as learning rate, activation function, etc.) through grid search or random search. In addition, regularization can be introduced (by adding regularization terms to the loss function, such as L1 or L2 regularization) to reduce the risk of model overfitting and improve the generalization ability of the model.

[0088] By using the abnormality detection model obtained through the above training, it is possible to detect in real time whether the crane operation state is abnormal. The crane in the embodiment of the present invention may be a crane used in various fields and environments, such as a crane used in a chemical enterprise.

[0089] like Figure 3 FIG. 1 is a flow chart of a method for detecting abnormal operation status of a crane pipe according to an embodiment of the present invention, comprising the following steps: Step 301, during the operation of the crane, the operation parameters of the crane are collected in real time.

[0090] The operating parameters may specifically include, but are not limited to, temperature, flow rate, pressure and other data.

[0091] Step 302: Fusing the operating parameters to obtain fused data.

[0092] The parameter fusion method can adopt weighted calculation. The weight determination of each parameter can refer to the parameter weight method in the previous anomaly detection model training process, which will not be repeated here.

[0093] Step 303: input the fused data into a pre-trained anomaly detection model, and obtain a reconstruction error according to the output of the anomaly detection model.

[0094] The anomaly detection model uses an autoencoder model, whose output is the reconstruction error of the corresponding input data, that is, the difference between the original input data and the model-reconstructed data.

[0095] Since the autoencoder model is trained on normal operating data, it learns to reconstruct the features of the normal state. Since the anomaly detection model does not directly learn the abnormal data, the reconstruction error is low for normal data but high for abnormal data.

[0096] Step 304: determine whether the crane operation status is abnormal according to the reconstruction error.

[0097] Specifically, a reconstruction error threshold may be set. If the reconstruction error currently detected is greater than the threshold, it is determined that the crane operation state is abnormal; otherwise, it is determined that the crane operation state is normal.

[0098] In some embodiments, the threshold value may be set based on empirical values.

[0099] In some embodiments, considering the changes in factors such as the crane operation environment, the threshold value may also be dynamically determined by statistically analyzing the distribution of reconstruction errors, as follows: Use the reconstruction error of normal operation data to calculate the distribution of reconstruction error (e.g. mean and standard deviation Specifically, the distribution of reconstruction errors obtained within a certain period of time can be statistically analyzed, and the reconstruction error threshold can be determined based on the statistical reconstruction error distribution. ,For example:

[0100] in, It is a sensitivity control parameter, which is usually set to 2 or 3. A higher value can reduce the false positive rate but may reduce sensitivity to anomalies.

[0101] The crane operation state abnormality detection method provided by the present invention collects crane operation parameters in real time during the crane operation process, calculates the reconstruction error of the current operation using a pre-built abnormality detection model, and determines whether the crane operation state is abnormal based on the reconstruction error. The abnormality detection model is based on the unsupervised learning ability of the autoencoder model to model the normal mode of crane operation, which is simple and convenient to implement, can effectively process multi-dimensional and highly complex data, and improve the accuracy and reliability of abnormality detection of crane operation state.

[0102] Furthermore, in some embodiments, when an abnormal operating status of the crane pipe is detected, an alarm may be triggered (for example, relevant operators may be notified through sound and light alarms, text messages, or cloud platform push notifications) to ensure a timely response; at the same time, the abnormal information (for example, timestamps, sensor values, reconstruction errors, etc.) may be recorded in a log file or database to provide corresponding data support for subsequent problem troubleshooting and analysis.

[0103] Accordingly, an embodiment of the present invention further provides a device for detecting abnormal operation status of a crane pipe, such as Figure 4 Shown is a structural schematic diagram of the device.

[0104] In this embodiment, the crane pipe operation state abnormality detection device 400 is pre-installed with an abnormality detection model 50, and the crane pipe operation state abnormality detection device 400 includes the following modules: The data acquisition module 401 is used to collect the operation parameters of the crane in real time during the operation of the crane; A data fusion module 402 is used to fuse the operating parameters to obtain fused data; A reconstruction error determination module 403 is used to input the fused data into a pre-trained autoencoder-based anomaly detection model, and obtain a reconstruction error according to the output of the anomaly detection model 50; The judgment module 404 is used to determine whether the operation status of the crane is abnormal according to the reconstruction error.

[0105] The anomaly detection model can be obtained by pre-training the corresponding model construction module based on the time series working parameter sequence collected under the normal operating state of the crane pipe. The model construction module can be a part of the device of the present invention or independent of the device, and this embodiment of the present invention is not limited to this.

[0106] In some embodiments, the crane pipe operation status abnormality detection device 400 may further include: an abnormality handling module (not shown in the figure), which is used to trigger an alarm when an abnormal crane pipe operation status is detected, and record the abnormal information in a log file or database.

[0107] like Figure 5 As shown, it is a structural schematic diagram of a model building module according to an embodiment of the present invention.

[0108] The model building module 500 includes the following units: The data acquisition unit 501 is used to acquire the working parameter sequence of the crane pipe based on the time series; A preprocessing unit 502, used for cleaning the working parameter sequence to obtain a cleaned data sequence; A normalization processing unit 503, used for performing normalization processing on each working parameter in the data sequence; A fusion unit 504 is used to fuse multiple heterogeneous data at each time point in the data sequence to obtain a fused data sequence; A network structure design unit 505, used to design a network structure of an anomaly detection model based on an autoencoder and determine a loss function; The training unit 506 is configured to use the fused data sequence as training data and train model parameters based on the loss function.

[0109] The specific implementation of each module and unit in the above embodiment can refer to the description in the above method embodiment of the present invention.

[0110] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0111] The present invention also provides a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and the computer program can be executed when it is run. Figure 1 or Figure 3 The storage medium may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc. The storage medium may also include a non-volatile memory (non-volatile) or a non-transitory memory, etc.

[0112] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, a computer, a server or a data provider to another website site, a computer, a server or a data provider by wired or wireless means.

[0113] The embodiments of the present invention are described in detail above. The present invention is described in detail using specific implementation methods herein. The description of the above embodiments is only used to help understand the method and system of the present invention. It is only a part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should belong to the scope of protection of the present invention, and the content of this specification should not be understood as limiting the present invention. Therefore, any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for detecting abnormal operation status of a crane pipe, characterized in that: The method comprises: During the operation of the crane pipe, the crane pipe operation parameters are collected in real time; fusing the operating parameters to obtain fused data; Inputting the fused data into a pre-trained anomaly detection model, and obtaining reconstructed data according to an output of the anomaly detection model; Calculating a reconstruction error based on the reconstructed data and the corresponding fused data; Whether the crane operation state is abnormal is determined according to the reconstruction error.

2. The method for detecting abnormal operation status of a crane pipe according to claim 1, characterized in that: The process of training an autoencoder-based anomaly detection model includes: Collect the working parameter sequence of the crane pipe based on the time series; Cleaning the working parameter sequence to obtain a cleaned data sequence; Normalizing each operating parameter in the data sequence; fusing a plurality of heterogeneous data at each time point in the data sequence to obtain a fused data sequence; Design the network structure of the autoencoder-based anomaly detection model and determine the loss function; The fused data sequence is used as training data, and model parameters are trained based on the loss function.

3. The method for detecting abnormal operation status of a crane pipe according to claim 2, characterized in that: The working parameters include temperature, flow rate, and pressure data.

4. The method for detecting abnormal operation status of a crane pipe according to claim 2, characterized in that: The cleaning of the working parameter sequence includes any one or more of the following: Processing abnormal values ​​and missing values ​​in the working parameter sequence; A noise reduction process is performed on the working parameter sequence.

5. The method for detecting abnormal operation status of a crane pipe according to claim 2, characterized in that: The fusing of the plurality of heterogeneous data at each time point in the data sequence comprises: Determining the weight of each operating parameter in the data sequence; A weighted fusion value of multiple heterogeneous data at each time point in the data sequence is calculated according to the weight of each working parameter.

6. The method for detecting abnormal operation status of a crane pipe according to claim 5, characterized in that: Determining the weight of each operating parameter in the data sequence includes: The weight of each working parameter in the data sequence is determined based on the following principles: (1) Dynamically adjust weights based on data reliability; (2) Assign weights based on the sensitivity of sensors to changes in variables.

7. The method for detecting abnormal operation status of a crane pipe according to claim 2, characterized in that: The autoencoder includes: an encoder and a decoder; the encoder includes a series of fully connected layers or convolutional layers that are reduced layer by layer; the decoder includes a layer-by-layer expansion network that is symmetrical to the encoder.

8. The method for detecting abnormal operation status of a crane pipe according to claim 2, characterized in that: The using the fused data sequence as training data and training the model parameters based on the loss function further comprises: Dividing the fused data sequence into a training set and a test set; Training model parameters using the training set and the loss function; The test set is used to evaluate the model effect, and the network structure and parameters of the model are optimized based on the evaluation results.

9. The method for detecting abnormal operation status of a crane pipe according to claim 1, characterized in that: Determining whether the crane operation state is abnormal according to the reconstruction error includes: If the reconstruction error is greater than a set threshold, it is determined that the crane operation state is abnormal; Otherwise, make sure the crane is operating normally.

10. The method for detecting abnormal operation status of a crane pipe according to claim 9, characterized in that: The method further comprises: According to the reconstruction error of normal operation data, the error distribution is statistically analyzed; The set threshold is determined according to the error distribution.

11. The method for detecting abnormal operation status of a crane pipe according to any one of claims 1 to 10, characterized in that: The method further comprises: When an abnormal operation status of the crane pipe is detected, an alarm is triggered and the abnormal information is recorded in a log file or database.

12. A crane pipe operation status abnormality detection device, characterized in that: The device comprises: Data acquisition module, used to collect crane operation parameters in real time during the operation of the crane; A data fusion module, used for fusing the operating parameters to obtain fused data; A reconstruction error determination module, used for inputting the fused data into a pre-trained autoencoder-based anomaly detection model, and obtaining a reconstruction error according to an output of the anomaly detection model; The judgment module is used to determine whether the crane operation state is abnormal according to the reconstruction error.

13. The crane pipe operation status abnormality detection device according to claim 12, characterized in that: The device also includes: a model building module for training an anomaly detection model based on an autoencoder; The model building module includes: A data acquisition unit, used for acquiring a time series-based working parameter sequence of the crane pipe; A preprocessing unit, used for cleaning the working parameter sequence to obtain a cleaned data sequence; A normalization processing unit, used for performing normalization processing on each working parameter in the data sequence; A fusion unit, used for fusing a plurality of heterogeneous data at each time point in the data sequence to obtain a fused data sequence; The network structure design unit is used to design the network structure of the autoencoder-based anomaly detection model and determine the loss function; A training unit is used to use the fused data sequence as training data and train model parameters based on the loss function.

14. The crane pipe operation status abnormality detection device according to claim 12 or 13, characterized in that: The device also includes: The exception handling module is used to trigger an alarm when an abnormal operation status of the crane pipe is detected, and record the abnormal information in a log file or database.

15. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for detecting abnormal operation status of a crane pipe according to any one of claims 1 to 11 are executed.

Images