Fault prediction model, method, equipment and medium for cooling device of CNC machine tool

Abstract

The present invention discloses a fault prediction model, method, device and medium for a cooling device of a numerically controlled machine tool, including: an input feature extractor for extracting spatial features of operation data of the cooling device of a numerically controlled machine tool; a data fusion unit for fusing the operation data and the spatial features to obtain the current hidden state; a dynamic feature analyzer for estimating the evolution trend of the operation data according to the hidden state, a comprehensive feature integrator for adaptively fusing the results of different features output by the dynamic feature analyzer, and a final prediction output generator for converting the fused hidden state into a specific prediction output. The probability of failure of a numerically controlled machine tool can be predicted, thereby achieving preventive maintenance and improving production efficiency and equipment service life.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault prediction, and in particular relates to a fault prediction model, method, equipment and medium for a cooling device of a numerically controlled machine tool. Background Art

[0002] CNC machine tools (CNC) have been widely used in modern manufacturing. Their high precision, high efficiency and automation make them important equipment for various processing operations. The cooling device is one of the key auxiliary equipment in CNC machine tools. It is mainly used to control the temperature of the machine tools, lubricate the tools and remove chips during processing. The cooling device in CNC machine tools plays an important role in controlling temperature, lubricating tools and removing chips, which directly affects the processing accuracy and service life of the machine tools. Common cooling devices include immersion cooling, spray cooling and atomization cooling, which effectively dissipate heat and lubricate in different ways. The failure of the cooling device will affect the processing quality and the stability of equipment operation. Therefore, it is necessary to predict the failure of the cooling device of CNC machine tools in order to predict the probability of failure of CNC machine tools, and then take corresponding measures to improve production efficiency and equipment service life. Summary of the invention

[0003] The object of the present invention is to provide a fault prediction model, method, device and storage medium for a cooling device of a CNC machine tool, which can predict the probability of failure of the CNC machine tool, thereby realizing preventive maintenance and improving production efficiency and equipment service life.

[0004] The first aspect of the present invention discloses a fault prediction model for a cooling device of a numerically controlled machine tool, comprising:

[0005] An input feature extractor is used to extract spatial features of operation data of a cooling device of a CNC machine tool;

[0006] A data fusion unit, used for fusing the operation data and the spatial features to obtain a current hidden state;

[0007] A dynamic feature analyzer is used to estimate the evolution trend of the operation data, and the dynamic feature analyzer includes a short-term change capturer, a long-term pattern detector and a spatial flow detector; the short-term change capturer is used to calculate the difference between the current hidden state and the hidden state at the previous moment to obtain a first result; the long-term pattern detector is used to obtain candidate states and important information at the current moment based on all historical hidden states, and obtain a second result based on the candidate states and the important information; the spatial flow detector is used to capture the local differences of the operation data and the information flow in space to obtain a third result;

[0008] A comprehensive feature integrator, used to fuse the first result, the second result and the third result to obtain a comprehensive feature;

[0009] The prediction output generator is used to obtain and output the prediction result based on the comprehensive features and the current hidden state.

[0010] In some embodiments, a time series segmentation unit is further included, and the time series segmentation unit is used to perform time series segmentation on the operation data to convert the operation data into time series data.

[0011] In some embodiments, the input feature extractor includes a first extraction unit and a second extraction unit, the first extraction unit is used to use a one-dimensional convolutional neural network to perform feature extraction on the time series data to obtain a first extraction result; the second extraction unit is used to use a one-dimensional convolutional neural network to perform feature extraction on the first extraction result to obtain the spatial feature.

[0012] In some embodiments, the data fusion unit includes a connection unit and a convolution unit, the connection unit is used to connect the operating data and the spatial features to obtain an input sequence; the convolution unit is used to use a one-dimensional convolution layer to extract features from the input sequence to obtain a current hidden state.

[0013] In some embodiments, the long-term pattern detector includes a candidate state unit, an important information unit, a final state unit and a convolution unit, wherein the candidate state unit is used to calculate the candidate state based on the difference between the current hidden state and the hidden state at the previous moment and the current hidden state; the important information unit is used to extract important information from the historical hidden state using an attention mechanism; the final state unit is used to calculate the final state based on the candidate state and the important information using a gating mechanism; and the convolution unit is used to perform a one-dimensional convolution operation on the final state to obtain the second result.

[0014] In some embodiments, the spatial flow detector includes a local difference unit and an information flow unit. The local difference unit is used to calculate the difference between the hidden state at the current moment and the hidden state of the neighboring position to obtain the local difference; the information flow unit is used to use the attention mechanism to calculate the attention weight according to the local difference, extract the flow feature according to the attention weight, and obtain the third result.

[0015] In some embodiments, the prediction output generator includes a splicing unit and a convolution unit, the splicing unit is used to splice the current hidden state and the comprehensive feature to obtain a splicing result; the convolution unit is used to perform a one-dimensional convolution operation on the splicing result to obtain the prediction result.

[0016] A second aspect of the present invention discloses a fault prediction method for a cooling device of a numerically controlled machine tool, comprising:

[0017] Obtain the operating data of the cooling device of CNC machine tools;

[0018] Preprocessing the operating data to obtain preprocessed data;

[0019] The preprocessed data is input into any of the above-mentioned fault prediction models to obtain the prediction results and output them.

[0020] The third aspect of the present invention discloses an electronic device, comprising a memory storing executable program code and a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the fault prediction method for a CNC machine tool cooling device disclosed in the second aspect.

[0021] A fourth aspect of the present invention discloses a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the fault prediction method for a cooling device of a CNC machine tool disclosed in the second aspect.

[0022] The beneficial effect of the present invention is that the spatial features of the operation data of the cooling device of the numerical control machine tool are extracted by the input feature extractor, the data fusion unit fuses the spatial features with the operation data to update the hidden state, and then the short-term change capturer, the long-term pattern detector and the spatial flow detector in the dynamic feature analyzer estimate different evolution trends according to the hidden state, and then the comprehensive feature integrator adaptively fuses various trends and predicts through the prediction output generator to obtain the prediction result, which can identify the signs of faults. Through effective fault prediction, preventive maintenance can be achieved, and production efficiency and equipment service life can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings herein show specific examples of the technical solutions described in the present invention, and together with the specific implementation methods, constitute a part of the specification, and are used to explain the technical solutions, principles and effects of the present invention.

[0024] Unless otherwise specified or defined, the same reference numerals in different drawings represent the same or similar technical features, and the same or similar technical features may also be represented by different reference numerals.

[0025] Figure 1 It is a module structure diagram of a fault prediction model of a cooling device of a numerically controlled machine tool disclosed in an embodiment of the present invention;

[0026] Figure 2 is a data flow diagram of a fault prediction model according to an embodiment of the present invention;

[0027] Figure 3 is a functional structure diagram of a long-term pattern detector in a fault prediction model according to an embodiment of the present invention;

[0028] Figure 4 is a functional structure diagram of a space flow detector in a fault prediction model of an embodiment of the present invention;

[0029] Figure 5 It is a fitting diagram of the predicted value and the true value of the fault prediction model of the cooling device of the numerical control machine tool according to the embodiment of the present invention on the test set;

[0030] Figure 6 It is a structural schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0031] Unless otherwise specified or defined, all technical and scientific terms used herein have the same meaning as those generally understood by those skilled in the art. In combination with the technical solution of the present invention in a realistic scenario, all technical and scientific terms used herein may also have meanings corresponding to the purpose of implementing the technical solution of the present invention. The "first, second..." used herein is only used to distinguish the names and does not represent a specific quantity or order. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0032] It should be noted that when a component is considered to be "fixed to" another component, it can be directly fixed to the other component or there can be a central component; when an component is considered to be "connected to" another component, it can be directly connected to the other component or there can be a central component at the same time; when an component is considered to be "installed on" another component, it can be directly installed on the other component or there can be a central component at the same time. When an component is considered to be "set on" another component, it can be directly set on the other component or there can be a central component at the same time.

[0033] Unless otherwise specified or defined, the "said" and "the" used in this document refer to the technical features or technical contents mentioned or described before the corresponding position, and the technical features or technical contents may be the same as or similar to the technical features or technical contents mentioned therein. In addition, the terms "including" and "having" and any variations thereof used in this document are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0034] In order to achieve preventive maintenance of the cooling device of CNC machine tools and improve production efficiency and equipment service life, the present invention proposes a fault prediction model for the cooling device of CNC machine tools, which is used to predict the probability of failure of the CNC machine tool in the future and take corresponding control measures according to the probability, thereby avoiding failure of the cooling device of the CNC machine tool.

[0035] like Figure 1 As shown, the fault prediction model of this embodiment is composed of an input feature extractor 610, a data fusion unit 620, a dynamic feature analyzer 630, a comprehensive feature integrator 640 and a prediction output generator 650. The input feature extractor 610 converts the operation data of the cooling device of the CNC machine tool into a high-dimensional representation of the hidden state and extracts the spatial features of the operation data; the data fusion unit 620 fuses the operation data and the spatial features and updates the current hidden state; the dynamic feature analyzer 630 includes a short-term change capturer, a long-term pattern detector and a spatial flow detector, which are used to estimate different evolution trends of the operation data; the comprehensive feature integrator 640 adaptively fuses the results of different features output by the dynamic feature analyzer; the prediction output generator 650 converts the fused hidden state into a specific prediction result and outputs it.

[0036] In this embodiment, data monitoring is performed on the operating equipment, and the key parameter monitored is the coolant temperature, that is, the operating data of the CNC machine tool cooling device is the temperature change data. Specifically, the temperature sensor is installed on the cooling device, and a preliminary calibration is performed to ensure the accuracy and reliability of the sensor. The temperature sensor is used to monitor the temperature change of the coolant, and the sensor is regularly checked and recalibrated to prevent data errors caused by sensor drift; the frequency of data collection is determined, for example: the collection frequency is set to perform temperature monitoring every 10 minutes to ensure that sufficient fault characteristics are captured and operating data is obtained.

[0037] Then the input feature extractor 610 extracts the features of the operation data. The input feature extractor 610 extracts spatial features from the original input data, maps the low-dimensional input data to a high-dimensional hidden state space, and can better represent the complexity of the input data. The original input data is preferably time series data. Therefore, the fault prediction model of this embodiment also includes a preprocessing unit and a time series segmentation unit. The preprocessing unit is used to preprocess the operation data of the CNC machine tool cooling device, and the time series segmentation unit is used to perform time series segmentation on the preprocessed operation data to convert the operation data into time series data.

[0038] Specifically, the preprocessing unit includes: data cleaning operation and data normalization operation. The data cleaning operation is to delete the sample points with a small number of randomly distributed missing values ​​in the data set, so as to improve the quality of the temperature change data set of the CNC machine tool cooling device; in order to accelerate convergence, make the gradient change smoother, and avoid numerical instability, the data is normalized. Among them, the normalization method adopts Standardization converts the cooling device temperature change data into a normal distribution with a mean of 0 and a standard deviation of 1. The normalization formula is:

[0039]

[0040] In the formula, is the original data value in the data set, is the mean of the data set, is the standard deviation of the data set, is the standardized data value.

[0041] The time series segmentation unit performs time series segmentation on the normalized data set to obtain subsequence data (i.e., time series data). During the segmentation process, the sliding window Set to 20, sliding step Set to 4. Reference Figure 2 , starting time is The subsequence data of the moment is represented as .

[0042] It should be noted that the time series segmentation unit may also be used as a part of the preprocessing unit to form an overall preprocessing unit.

[0043] The input feature extractor 610 uses a first extraction unit and a second extraction unit to extract spatial features from the time series data. The first extraction unit is used to extract features from the time series data using a one-dimensional convolutional neural network to obtain a first extraction result; the second extraction unit is used to extract features from the first extraction result using a one-dimensional convolutional neural network to obtain spatial features.

[0044] Specifically, a two-layer one-dimensional convolutional neural network is used to extract features from time series data. The first layer of convolution operation is as follows:

[0045]

[0046] In the formula, The start time is In this embodiment, the subsequence data of the time The size is , where 20 is the sliding window , 1 is a single channel, is a one-dimensional convolution operation, is the weight matrix of the first layer convolution kernel, is the bias term, is the activation function, is the output of the first convolutional layer, that is, the first extraction result.

[0047] After the first convolutional layer, the output The shape and size of , The number of convolution kernels used for the first convolution operation is the output of the first convolution layer. It is calculated as the input of the second convolutional layer. The second convolution operation is as follows:

[0048]

[0049] In the formula, is the weight matrix of the first layer convolution kernel, is the bias term, is the output of the second convolutional layer; after the second convolutional layer, the spatial features are output The shape and size of , The number of convolution kernels used for the second layer of convolution operations.

[0050] The input feature extractor 610 plays a role in promoting univariate time series data to a high-dimensional feature space in the fault prediction model of the cooling device of a CNC machine tool, providing a rich feature basis for subsequent spatiotemporal prediction. The rich features include diversified feature dimensions, high-dimensional feature representations, and fine-grained feature information.

[0051] After obtaining the spatial features, the spatial features are combined with the operation data (in this embodiment, the time series is segmented) by the data fusion unit 620. The operating data of the following steps are ) to obtain the current hidden state The data fusion unit 620 includes a connection unit and a convolution unit, wherein the connection unit is used to connect the operation data and the spatial features to obtain an input sequence; and the convolution unit is used to extract features from the input sequence using a one-dimensional convolution layer to obtain a current hidden state.

[0052] Specifically, a one-dimensional convolutional layer is used to The spatial features output by the input feature extractor Fusion is performed to generate a new hidden state. The implementation process is as follows:

[0053]

[0054] In the formula, To concatenate the input data with spatial features into an input sequence, is the weight matrix of the convolution kernel in the data fusion unit, is the bias term in the data fusion unit, is the output of the data fusion unit, that is, the current hidden state.

[0055] The data fusion unit 620 fuses the latest operating data with the extracted spatial features, which helps to integrate the new observation information into the model's memory to more accurately reflect the current status.

[0056] The dynamic feature analyzer 630 is used to estimate different evolution trends of the operation data. The dynamic feature analyzer 630 includes a short-term change capturer, a long-term pattern detector, and a spatial flow detector.

[0057] Among them, the short-term change capturer is more suitable for the relatively stable data of the operating temperature change of the cooling device of the CNC machine tool in the present invention, and can capture short-term linear changes. The output of the short-term change capturer is obtained by calculating the difference between the current hidden state and the hidden state at the previous moment. The formula of the short-term change capturer is as follows:

[0058]

[0059] In the formula, is the current hidden state, is the hidden state at the previous moment, is the output of the short-term change capturer, i.e. the first result, which reflects the change of the hidden state between two consecutive time steps.

[0060] The short-term change catcher captures local dynamic changes in time series data, enhances the model's responsiveness to short-term fluctuations, helps improve forecasting accuracy, and enables the model to adapt more flexibly to diverse time series patterns, thereby improving overall forecasting performance.

[0061] The long-term pattern detector captures long-term nonlinear dynamic changes by utilizing the global historical hidden state. The long-term pattern detector includes a candidate state unit, an important information unit, a final state unit, and a convolution unit. The candidate state unit is used to calculate the candidate state based on the difference between the current hidden state and the hidden state at the previous moment and the current hidden state; the important information unit is used to extract important information from the historical hidden state using an attention mechanism; the final state unit is used to calculate the final state based on the candidate state and the important information using a gating mechanism; the convolution unit is used to perform a one-dimensional convolution operation on the final state to obtain a second result.

[0062] Specifically, the long-term pattern detector uses the attention mechanism to weight all historical hidden states, extract the most important information for the current moment, and capture the trend in the time series through the attention mechanism and the gating mechanism, thereby providing a more comprehensive trend estimation; Figure 3 As shown, the implementation of the long-term pattern detector is as follows:

[0063] will contain the time from time point 1 to time point All hidden states of At the time point The hidden state of As the input of the long-term pattern detector; first calculate the hidden state change at each time step, using the following formula:

[0064]

[0065] In the formula, is the time step The difference between the hidden state of the next time step and the hidden state of the next time step; then the attention mechanism is used to calculate the trend of the current moment. The formula used is as follows:

[0066]

[0067]

[0068] In formula (1), for The hidden state at time step The difference in hidden state at each time step, For use with The multiplied weight matrix, For use with The multiplied weight matrix, for Activation function, for Reset the output of the gate at all times; in formula (2), For use with The multiplied weight matrix, For use with The multiplied weight matrix, is the element-wise product operation, is the activation function, The candidate state combines the state change at the previous moment and the historical state for trend estimation at the current moment;

[0069] Then extract the important information. The formula used to extract the information is as follows:

[0070]

[0071]

[0072]

[0073]

[0074] In the formula, is the weight matrix used to generate the query vector, is the weight matrix used to generate the key vector, is the attention weight, through Function Compute Query Vector With key vector The similarity of It is the result of weighted summation, which represents the important information extracted from the historical state;

[0075] The gating variables are then used to calculate the final state in the long-term pattern detector module that combines the candidate states and the important information. , the formula is as follows:

[0076]

[0077]

[0078] In formula (7), is a gating variable used to control the candidate state and important information The weighted ratio of To change the state of the previous moment The weight matrix mapped to the gating variable space, For the candidate state The weight matrix mapped to the gating variable space, For important information The weight matrix mapped to the gating variable space;

[0079] Finally, the final state Perform convolution operation to convert the state Converted to the output of the long-term pattern detector, the calculation formula is as follows:

[0080]

[0081] In formula (9), is a one-dimensional convolution operation, is the output of the long-term pattern detector, i.e., the second result.

[0082] The long-term pattern detector captures the overall trend changes in the time series, helps the model understand and predict the long-term trend, can effectively smooth the noise, and extract the global features in the data, thereby enhancing the model's ability to capture long-term trends and improving the stability and accuracy of the prediction.

[0083] The spatial flow detector captures local differences in time series data and spatial information flow, including local difference units and information flow units. The local difference unit is used to calculate the difference between the hidden state at the current moment and the hidden state of the neighboring position to obtain the local difference; the information flow unit is used to calculate the attention weight according to the local difference using the attention mechanism, and extract the flow features according to the attention weight.

[0084] Specifically, Figure 4 As shown, first calculate the local difference, which is the difference between the hidden state of the current moment and its neighborhood, and is used to capture local changes. The operation formula is as follows:

[0085]

[0086] In formula (10), Neighborhood location The hidden state of Central location The hidden state of is the local difference between the hidden states of the neighborhood and the center;

[0087] Then the attention mechanism is applied on the local differences to extract flow features. The steps are as follows:

[0088]

[0089]

[0090] In formula (11), For The transpose of the multiplied weight matrix, for The output of the spatial flow detector at time is the query matrix in the spatial flow detector module; in formula (12), For The transpose of the multiplied weight matrix, Neighborhood location The key matrix represents the features of the position and is used to compare with the query matrix Perform a dot product operation to generate attention weights , The calculation process is as follows:

[0091]

[0092] In formula (13), Neighborhood location The transpose of the query matrix; then the attention weights Apply to local differences Calculate the flow characteristics , the formula is as follows:

[0093]

[0094] is the weighted difference between the neighborhood position and the center position, reflecting the flow trend in the local area; finally, the flow characteristics captured by the spatial flow detector are updated, and the update formula is as follows:

[0095]

[0096] In formula (15), is the update gate, which is used to control the fusion of the flow trend at the previous moment and the flow characteristics at the current moment. is an element-by-element multiplication operation, For space flow detectors Capturing the flow trend at all times is the third result.

[0097] The spatial flow detector captures the spatial and directional changes in time series data, reflects the natural flow characteristics of information in the sequence, and enhances the model's understanding of spatiotemporal information by simulating dynamic processes similar to physical flow in the data, thereby improving the accuracy of prediction results and the ability to capture complex trends.

[0098] After obtaining the first result, the second result and the third result output by the dynamic feature analyzer, they are fused using the comprehensive feature integrator 640 to obtain the comprehensive feature.

[0099] Specifically, the comprehensive feature integrator 640 can adaptively fuse the results of different features to provide comprehensive features. The above comprehensive features can more comprehensively reflect the changing trend of time series data, thereby improving the accuracy of prediction. The feature fusion formula is as follows:

[0100]

[0101] In formula (16), is the output of the short-term change capturer, i.e., the first result; is the output of the long-term pattern detector, i.e., the second result; For space flow detectors The flow trend captured at every moment is the third result; , , are the coefficients of each term, and the sum is 1. It is the comprehensive feature, that is, the fused feature information.

[0102] The comprehensive feature integrator 640 fuses feature information at different levels, improving the model's ability to understand and model multi-scale trends, thereby enhancing the accuracy and robustness of predictions and enabling the model to more comprehensively capture complex patterns in the data.

[0103] The prediction output generator 650 converts the fused feature information into the final prediction value, including a splicing unit and a convolution unit. The splicing unit is used to splice the current hidden state and the comprehensive feature to obtain the splicing result; the convolution unit is used to perform a one-dimensional convolution operation on the splicing result to obtain the prediction result.

[0104] Specifically, firstly, the output of the data fusion unit Output of the comprehensive feature integrator Concatenate and then perform convolution to generate the final prediction value , the implementation formula is as follows:

[0105]

[0106] In the formula, is the bias term, It is a one-dimensional convolution operation.

[0107] After the fault prediction model is built, it needs to be trained before use. This embodiment collects the temperature change data of the cooling device of the CNC machine tool, performs data cleaning, normalization, and time series segmentation on the temperature change data of the cooling device, and then divides the preprocessed data into a training set and a test set. Among them, the normalized data set is ,in, is the total number of samples in the dataset, the total number of samples is 4000. During the time series data segmentation process, the sliding window Set to 20, sliding step Set to 4, the subsequences after time series segmentation using sliding windows are , a total of 995 time series, that is, 995 training samples; the 995 samples are divided into training set and test set in a ratio of 7:3. The training set is used to train the fault prediction model of the cooling device of the CNC machine tool. The training process mainly includes hyperparameter setting and back propagation process; the test set is used for fault prediction during testing.

[0108] Among them, the back propagation process uses the L1 loss function, the formula is as follows:

[0109]

[0110] In the formula, For the True labels of samples, For the The predicted value of the fault prediction model of the cooling device of a CNC machine tool, m is the number of batch samples, which is the same as The parameters are the same, is the loss value of the true value prediction value; the above hyperparameter settings include: Set to 512, The learning rate is 0.0002, The number of iterations is 1500 and the optimizer is .

[0111] Figure 5 This is the fitting of the predicted value and the true value of the fault prediction model of the cooling device of the CNC machine tool on the test set. Figure 5 It can be seen that with the increase of time, the true value is gradually increasing, and the predicted value of the fault prediction model is also increasing. At 60 minutes, the true value exceeds the warning line, and the prediction model also makes a correct prediction, which proves that the model can predict the future failure of the cooling device of the CNC machine tool.

[0112] In summary, the fault prediction model of this embodiment extracts the spatial features of the operation data of the cooling device of the CNC machine tool, integrates the spatial features and the operation data to update the hidden state, and then estimates different evolution trends according to the hidden state through the short-term change capturer, the long-term pattern detector and the spatial flow detector, and then adaptively integrates various trends and predicts to obtain prediction results, which can identify the signs of faults. Through effective fault prediction, preventive maintenance can be achieved, and production efficiency and equipment service life can be improved.

[0113] The embodiment of the present invention also discloses a method for predicting a fault of a cooling device of a numerically controlled machine tool, which is used to predict a fault of a cooling device of a numerically controlled machine tool. The method can be implemented by computer programming. The executor of the method can be an electronic device such as a computer, a laptop computer, a tablet computer, or a fault prediction device for a cooling device of a numerically controlled machine tool embedded in an electronic device, and the present invention does not limit this. First, the operating data of the cooling device of the numerically controlled machine tool is obtained, for example, a temperature sensor is used to monitor the temperature change of the coolant to obtain the operating data; then the operating data is preprocessed, for example: data cleaning, normalization, and time series segmentation, etc., to obtain the preprocessed data; the preprocessed data is input into a trained fault prediction model to obtain a prediction result, and corresponding preventive measures are taken according to the prediction result to avoid equipment failure and improve production efficiency and equipment service life.

[0114] like Figure 6 As shown, an embodiment of the present invention discloses an electronic device, including a memory 401 storing executable program codes and a processor 402 coupled to the memory 401;

[0115] The processor 402 calls the executable program code stored in the memory 401 to execute the fault prediction method for the cooling device of a numerically controlled machine tool described in the above embodiment.

[0116] An embodiment of the present invention further discloses a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the fault prediction method for a cooling device of a numerically controlled machine tool described in the above embodiment.

[0117] The purpose of the above embodiments is to exemplarily reproduce and deduce the technical solution of the present invention, and to fully describe the technical solution, purpose and effect of the present invention. Its purpose is to make the public understand the disclosed content of the present invention more thoroughly and comprehensively, and it does not limit the scope of protection of the present invention.

[0118] The above embodiments are not exhaustive enumerations of the present invention, and there may be multiple other implementations not listed. Any replacement and improvement made without violating the concept of the present invention shall fall within the protection scope of the present invention.

Claims

1. A fault prediction model for a cooling device of a CNC machine tool, characterized in that: include: An input feature extractor is used to extract spatial features of operation data of a cooling device of a CNC machine tool; A data fusion unit, used for fusing the operation data and the spatial features to obtain a current hidden state; A dynamic feature analyzer, used for estimating the evolution trend of the operation data, the dynamic feature analyzer comprising a short-term change capturer, a long-term pattern detector and a spatial flow detector; The short-term change capturer is used to calculate the difference between the current hidden state and the hidden state at the previous moment to obtain a first result; the long-term pattern detector is used to obtain candidate states and important information at the current moment based on all historical hidden states, and obtain a second result based on the candidate states and the important information; the spatial flow detector is used to capture the local differences of the running data and the information flow in space to obtain a third result; A comprehensive feature integrator, used to fuse the first result, the second result and the third result to obtain a comprehensive feature; A prediction output generator, used to obtain and output a prediction result based on the comprehensive features and the current hidden state; The long-term pattern detector includes a candidate state unit, an important information unit, a final state unit and a convolution unit, wherein the candidate state unit is used to calculate the candidate state according to the difference between the current hidden state and the hidden state at the previous moment and the current hidden state; the important information unit is used to extract important information from the historical hidden state by using an attention mechanism; the final state unit is used to calculate the final state according to the candidate state and the important information by using a gating mechanism; the convolution unit is used to perform a one-dimensional convolution operation on the final state to obtain the second result; The spatial flow detector includes a local difference unit and an information flow unit. The local difference unit is used to calculate the difference between the hidden state at the current moment and the hidden state of the neighboring position to obtain the local difference; the information flow unit is used to adopt an attention mechanism to calculate the attention weight according to the local difference, extract the flow feature according to the attention weight, and obtain the third result.

2. The fault prediction model according to claim 1, characterized in that: It also includes a time series segmentation unit, which is used to perform time series segmentation on the operating data and convert the operating data into time series data.

3. The fault prediction model according to claim 2, characterized in that: The input feature extractor includes a first extraction unit and a second extraction unit. The first extraction unit is used to use a one-dimensional convolutional neural network to perform feature extraction on the time series data to obtain a first extraction result; the second extraction unit is used to use a one-dimensional convolutional neural network to perform feature extraction on the first extraction result to obtain the spatial feature.

4. The fault prediction model according to claim 1, characterized in that: The data fusion unit includes a connection unit and a convolution unit. The connection unit is used to connect the operating data and the spatial features to obtain an input sequence; the convolution unit is used to extract features from the input sequence using a one-dimensional convolution layer to obtain a current hidden state.

5. The fault prediction model according to claim 1, characterized in that: The prediction output generator includes a splicing unit and a convolution unit. The splicing unit is used to splice the current hidden state and the comprehensive feature to obtain a splicing result; the convolution unit is used to perform a one-dimensional convolution operation on the splicing result to obtain the prediction result.

6. A method for predicting a fault of a cooling device of a numerically controlled machine tool, characterized in that: include: Obtain the operating data of the cooling device of CNC machine tools; Preprocessing the operating data to obtain preprocessed data; The preprocessed data is input into the fault prediction model as described in any one of claims 1 to 5 to obtain and output the prediction results.

7. An electronic device, characterized in that: It comprises a memory storing executable program code and a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the fault prediction method for the cooling device of a CNC machine tool as claimed in claim 6.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program enables a computer to execute the fault prediction method for a cooling device of a numerically controlled machine tool according to claim 6.

Images