Traction machine main shaft module life prediction method, system, medium, equipment and product based on deep convolutional neural network

Abstract

The present invention provides a method, system, medium, equipment and product for predicting the life of a traction machine main shaft module based on a deep convolutional neural network, and relates to the field of life prediction of a traction machine main shaft module. The life prediction method comprises: obtaining heat flux data of a main shaft bearing in the traction machine main shaft module rotating a predetermined number of times; generating a heat flux change curve according to the heat flux data; generating a time-frequency chromatogram according to the heat flux change curve; inputting the time-frequency chromatogram into a life prediction model for the traction machine main shaft module to obtain the predicted life data of the traction machine main shaft module. The present invention can be used to effectively predict the operating state and life of the traction machine main shaft module through the fluctuation of the heat flux value of the main shaft bearing, so as to ensure the reliability and safety of the elevator operation.

Description

Technical Field

[0001] The present invention relates to the field of life prediction of a main shaft module of a traction machine, and in particular, to a method, system, medium, equipment and product for predicting the life of a main shaft module of a traction machine based on a deep convolutional neural network. Background Art

[0002] The traction machine main shaft module is an elevator transmission component composed of the traction machine main shaft, main shaft bearing, traction wheel and related connecting parts. During use, the various parts of the traction machine main shaft module will be subjected to variable loads, especially the traction machine main shaft, which is prone to deformation and wear during use, and has a great impact on the functions of other connected parts. In severe cases, it may cause the elevator lifting transmission to fail and affect the safe operation of the elevator. Therefore, in order to ensure the safe and reliable operation of the elevator, it is necessary to effectively predict the life of the traction machine main shaft module before it fails.

[0003] However, the current analysis and prediction of the life of the traction machine main shaft module is mainly carried out by shutting down the elevator, disassembling the traction machine main shaft module on site, and then testing the main components, analyzing the functional status of the traction machine main shaft, bearings and traction wheels, and evaluating the overall life of the module. In particular, in the detection of the traction machine main shaft, all the connecting parts on it need to be disassembled for detection, and the whole process is relatively complicated. In addition, in addition to being able to perform on-site inspections after disassembly, the detection of small parts such as bearings in the traction machine main shaft module can also predict the life of the bearings through bearing vibration analysis and friction heat analysis.

[0004] It can be seen that the existing prediction methods for the life of the traction machine main shaft module are relatively complicated, and it is necessary to disassemble the main parts of the traction machine main shaft module for detection and analysis, which greatly reduces the detection efficiency and cannot grasp the operating status of the traction machine main shaft in real time. In addition, although the detection of small parts such as bearings in the module can be predicted and analyzed by friction heat analysis, it is only an isolated prediction and analysis for a single part, and there is no effective connection with the operating status of the main shaft. It is impossible to predict the life of the traction machine main shaft module from the mutual connection relationship between the main parts (such as the traction machine main shaft and the main shaft bearing) and the functional relationship. Summary of the invention

[0005] In order to effectively predict the operating life of the entire hoisting machine main shaft module, an embodiment of the present invention provides a method, system, device, storage medium and computer program product for predicting the life of a hoisting machine main shaft module based on a deep convolutional neural network. The heat flux of the bearings of the hoisting machine main shaft module under different operating conditions is analyzed and identified through the hoisting machine main shaft module life prediction model to determine the intrinsic connection between the operating state of the main shaft bearing and the deformation state of the traction machine main shaft, thereby being able to effectively predict the life of the traction machine main shaft module according to the fluctuation of the heat flux value of the main shaft bearing.

[0006] According to the first aspect of the present invention, an embodiment of the present invention provides a method for predicting the life of a hoisting machine main shaft module based on a deep convolutional neural network, the method comprising: obtaining heat flux data of a main shaft bearing in the traction machine main shaft module rotating a predetermined number of times; generating a heat flux change curve based on the heat flux data; generating a time-frequency chromatogram based on the heat flux change curve; inputting the time-frequency chromatogram into a life prediction model for the traction machine main shaft module to obtain predicted life data of the traction machine main shaft module.

[0007] The above-mentioned embodiment of the present invention analyzes and identifies the heat flux of the bearings of the traction machine main shaft module under different operating conditions through the life prediction model of the traction machine main shaft module, thereby determining the intrinsic connection between the operating state of the main shaft bearing and the deformation state of the traction machine main shaft. It can effectively predict the operating state and life of the traction machine main shaft module through the fluctuation of the heat flux value of the main shaft bearing, so as to ensure the reliability and safety of the elevator operation.

[0008] In some embodiments of the present invention, the life prediction model of the hoisting machine main shaft module is obtained in the following manner: according to the digital model of the hoisting machine main shaft module, a first heat flux transformation curve of the hoisting machine main shaft rotating the predetermined number of times under different states, a time-frequency chromatogram corresponding to the first heat flux transformation curve, and a main shaft bearing fatigue life value are obtained as a test set and a training set; according to the test set and the training set, the life prediction model of the hoisting machine main shaft module constructed based on the convolutional neural network is trained and evaluated, and the parameters of the life prediction model of the hoisting machine main shaft module are adjusted according to the evaluation results.

[0009] The above-mentioned embodiment of the present invention obtains relevant training data and test data through the digital model of the hoisting machine main shaft module, and obtains the life prediction model of the hoisting machine main shaft module based on the test set and training set obtained from the digital model of the hoisting machine main shaft module, which can improve the generalization performance of the prediction model, and then realize the accurate prediction of the life of the elevator traction machine main shaft module in different environments and types, thereby ensuring the safe operation of the elevator.

[0010] In some embodiments of the present invention, a digital model of the traction machine main shaft module is generated according to the structural parameters, operating parameters, and environmental parameters of the traction machine main shaft module.

[0011] In some embodiments of the present invention, a first change curve of the first heat flux of the main shaft bearing of the traction machine main shaft module under normal operating conditions is obtained through the digital model of the traction machine main shaft module; an actual change curve of the first heat flux of the main shaft bearing of the traction machine main shaft module under actual normal operating conditions is obtained in a database; and the digital model of the traction machine main shaft module is corrected according to the first change curve and the actual change curve.

[0012] In some embodiments of the present invention, a first heat flux transformation curve of the traction machine main shaft when the main shaft is normal and the bearings are worn, a time-frequency chromatogram corresponding to the first heat flux transformation curve, and a fatigue life value of the main shaft bearing are obtained according to the digital model of the traction machine main shaft module; a first heat flux transformation curve of the traction machine main shaft when the main shaft is deformed and worn and the bearings are normal, a time-frequency chromatogram corresponding to the first heat flux transformation curve, and a fatigue life value of the main shaft bearing are obtained according to the digital model of the traction machine main shaft module.

[0013] According to the second aspect of the present invention, an embodiment of the present invention provides a life prediction system for a hoisting machine main shaft module based on a deep convolutional neural network, and the life prediction system includes: a heat flux data acquisition module, used to obtain the heat flux data of the main shaft bearing in the traction machine main shaft module rotating a predetermined number of times; a time-frequency chromatogram generation module, used to generate a heat flux change curve according to the heat flux data, and generate a time-frequency chromatogram according to the heat flux change curve; a life prediction module, used to input the time-frequency chromatogram into the life prediction model of the traction machine main shaft module to obtain the predicted life data of the traction machine main shaft module.

[0014] The above-mentioned embodiment of the present invention analyzes and identifies the heat flux of the bearings of the traction machine main shaft module under different operating conditions through the life prediction model of the traction machine main shaft module, thereby determining the intrinsic connection between the operating state of the main shaft bearing and the deformation state of the traction machine main shaft. It can effectively predict the operating state and life of the traction machine main shaft module through the fluctuation of the heat flux value of the main shaft bearing, so as to ensure the reliability and safety of the elevator operation.

[0015] In some embodiments of the present invention, the life prediction system also includes a model acquisition module, which is used to obtain the life prediction model of the hoisting machine main shaft module in the following manner: according to the digital model of the hoisting machine main shaft module, the maximum heat flux transformation curve of the traction machine main shaft rotating one circle under different states, the time-frequency chromatogram corresponding to the maximum heat flux transformation curve, and the fatigue life value of the main shaft bearing are obtained as a test set and a training set; according to the test set and the training set, the life prediction model of the traction machine main shaft module constructed based on the convolutional neural network is trained and evaluated, and the parameters of the life prediction model of the traction machine main shaft module are adjusted according to the evaluation results.

[0016] The above-mentioned embodiment of the present invention obtains relevant training data and test data through the digital model of the hoisting machine main shaft module, and obtains the life prediction model of the hoisting machine main shaft module based on the test set and training set obtained from the digital model of the hoisting machine main shaft module, which can improve the generalization performance of the prediction model, and then realize the accurate prediction of the life of the elevator traction machine main shaft module in different environments and types, thereby ensuring the safe operation of the elevator.

[0017] In some embodiments of the present invention, a digital model of the traction machine main shaft module is generated according to the structural parameters, operating parameters, and environmental parameters of the traction machine main shaft module.

[0018] In some embodiments of the present invention, a first change curve of the first heat flux of the main shaft bearing of the traction machine main shaft module under normal operating conditions is obtained through the digital model of the traction machine main shaft module; an actual change curve of the first heat flux of the main shaft bearing of the traction machine main shaft module under actual normal operating conditions is obtained in a database; and the digital model of the traction machine main shaft module is corrected according to the first change curve and the actual change curve.

[0019] In some embodiments of the present invention, a first heat flux transformation curve of the traction machine main shaft when the main shaft is normal and the bearings are worn, a time-frequency chromatogram corresponding to the first heat flux transformation curve, and a fatigue life value of the main shaft bearing are obtained according to the digital model of the traction machine main shaft module; a first heat flux transformation curve of the traction machine main shaft when the main shaft is deformed and worn and the bearings are normal, a time-frequency chromatogram corresponding to the first heat flux transformation curve, and a fatigue life value of the main shaft bearing are obtained according to the digital model of the traction machine main shaft module.

[0020] According to the third aspect of the present invention, an embodiment of the present invention provides a computer-readable storage medium having computer-readable instructions stored thereon. When the computer-readable instructions are executed by a processor, the computer performs the following operations: the operations include the steps included in the method for predicting the life of a traction machine main shaft module based on a deep convolutional neural network as described in any of the above embodiments.

[0021] According to the fourth aspect of the present invention, an embodiment of the present invention provides a computer device including a memory and a processor, wherein the memory is used to store one or more computer-readable instructions, wherein when the one or more computer-readable instructions are executed by the processor, the method for predicting the life of a traction machine main shaft module based on a deep convolutional neural network as described in any one of the above embodiments can be implemented.

[0022] According to a fifth aspect of the present invention, an embodiment of the present invention provides a computer program product comprising a computer program, which, when executed by a processor, implements the method for predicting the life of a traction machine main shaft module based on a deep convolutional neural network as described in any of the above embodiments.

[0023] From the above, it can be seen that the method, system, device, medium and computer program product for predicting the life of the hoisting machine main shaft module based on the deep convolutional neural network provided in the embodiment of the present invention analyzes and identifies the heat flux of the bearings of the traction machine main shaft module under different operating conditions through the traction machine main shaft module life prediction model, thereby determining the intrinsic connection between the operating state of the main shaft bearing and the deformation state of the traction machine main shaft, and can effectively predict the operating state and life of the traction machine main shaft module through the fluctuation of the main shaft bearing heat flux value, so as to ensure the reliability and safety of the elevator operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic flow chart of a method for predicting the life of a main shaft module of a traction machine based on a deep convolutional neural network according to Embodiment 1 of the present invention;

[0025] Figure 2 It is a flow chart of a method for predicting the life of a main shaft module of a traction machine based on a deep convolutional neural network according to Embodiment 2 of the present invention;

[0026] Figure 3 It is a flow chart of a method for predicting the life of a main shaft module of a traction machine based on a deep convolutional neural network according to Embodiment 3 of the present invention;

[0027] Figure 4 It is an architectural diagram of a traction machine main shaft module life prediction system based on a deep convolutional neural network according to Example 4 of the present invention. DETAILED DESCRIPTION

[0028] The various aspects of the present invention are described in detail below in conjunction with the accompanying drawings and specific embodiments. Among them, well-known modules, units and their connections, links, communications or operations are not shown or described in detail. In addition, the described features, architectures or functions can be combined in any way in one or more embodiments. It should be understood by those skilled in the art that the various embodiments described below are only used for illustration and are not intended to limit the scope of protection of the present invention. It can also be easily understood that the modules or units or processing methods in each embodiment described herein and shown in the accompanying drawings can be combined and designed in various different configurations.

[0029] The following is a brief description of the terms used in this article.

[0030] Traction machine main shaft module: The traction machine main shaft module is a key component in the elevator traction system, which is mainly composed of a main shaft, bearings, couplings and other parts.

[0031] Heat flux: A physical quantity used to describe the rate of heat transfer through a unit area per unit time.

[0032] Time-frequency chromatogram: Time-frequency chromatogram, also known as time-frequency spectrum diagram or spectrogram, is a graph that represents a signal in two dimensions: time and frequency. It can show how a signal changes with time and frequency.

[0033] ReLU: Rectified Linear Unit, a widely used nonlinear activation function.

[0034] [Example 1]

[0035] Figure 1 It is a flow chart of a method for predicting the life of a main shaft module of a traction machine based on a deep convolutional neural network according to Example 1 of the present invention.

[0036] like Figure 1 As shown, in Embodiment 1 of the present invention, the life prediction method may include: step S11, step S12, step S13 and step S14, and the above steps are described in detail below.

[0037] In step S11, the heat flux data of the main shaft bearing in the traction machine main shaft module rotating a predetermined number of times is obtained. In this embodiment, the maximum heat flux data of the main shaft bearing in the traction machine main shaft module rotating one circle is obtained in step S11. The heat flux data is obtained by a heat flux sensor.

[0038] In step S12, a heat flux variation curve is generated according to the heat flux data.

[0039] In step S13, a time-frequency chromatogram is generated according to the heat flux variation curve. In this embodiment, the heat flux variation curve is converted into a corresponding time-frequency chromatogram by wavelet transformation.

[0040] In step S14, the time-frequency chromatogram is input into a life prediction model of a main shaft module of a traction machine to obtain predicted life data of the main shaft module of the traction machine.

[0041] In some embodiments, the life prediction model of the hoisting machine main shaft module is obtained in the following manner: according to the digital model of the hoisting machine main shaft module, a first heat flux transformation curve of the hoisting machine main shaft rotating the predetermined number of times under different states, a time-frequency chromatogram corresponding to the first heat flux transformation curve, and a main shaft bearing fatigue life value are obtained as a test set and a training set; according to the test set and the training set, the life prediction model of the hoisting machine main shaft module constructed based on the convolutional neural network is trained and evaluated, and the parameters of the life prediction model of the hoisting machine main shaft module are adjusted according to the evaluation results.

[0042] In this embodiment, obtaining the first heat flux transformation curve of the traction machine main shaft rotating the predetermined number of times under different states, the time-frequency chromatogram corresponding to the first heat flux transformation curve, and the fatigue life value of the main shaft bearing according to the digital model of the traction machine main shaft module includes: obtaining the first heat flux transformation curve of the traction machine main shaft rotating the predetermined number of times when the main shaft is normal and the bearing is worn according to the digital model of the traction machine main shaft module; and obtaining the first heat flux transformation curve of the traction machine main shaft rotating the predetermined number of times when the main shaft is deformed and worn and the bearing is normal according to the digital model of the traction machine main shaft module, the time-frequency chromatogram corresponding to the first heat flux transformation curve, and the fatigue life value of the main shaft bearing.

[0043] Wherein, the digital model of the traction machine main shaft module is generated according to the structural parameters, operating parameters and environmental parameters of the traction machine main shaft module, the operating parameters include but are not limited to: load range, starting speed, operating speed, etc., and the environmental parameters include but are not limited to: temperature and humidity of the space where the main shaft module is located, etc. Further, the digital model of the traction machine main shaft module is corrected in the following manner: obtaining the first change curve of the first heat flux of the main shaft bearing of the traction machine main shaft module under normal operating conditions through the digital model of the traction machine main shaft module; obtaining the actual change curve of the first heat flux of the main shaft bearing of the traction machine main shaft module under actual normal operating conditions in the database; and correcting the digital model of the traction machine main shaft module according to the first change curve and the actual change curve.

[0044] By adopting the above-mentioned life prediction method of Example 1 of the present invention, before the traction machine main shaft module fails, according to the influence of the main shaft deformation in the traction machine main shaft module on the operation of its bearings, by constructing a digital model of the traction machine main shaft module, the interaction relationship between the main shaft bearing and the traction machine main shaft is analyzed, and the heat flux of the bearing under different operating conditions of the traction machine main shaft module is analyzed and identified, so as to determine the intrinsic connection between the operating state of the main shaft bearing and the deformation state of the traction machine main shaft, so as to achieve effective prediction of the operating state and life of the traction machine main shaft module through the fluctuation of the heat flux value of the main shaft bearing, so as to ensure the reliability and safety of the elevator operation.

[0045] At the same time, the feasibility of inferring the deformation of the main shaft according to the change of heat flux is explained as follows:

[0046] First, there is a relationship between the heat generation of the roller bearing and the friction torque of the roller bearing as shown in the following formulas (1) and (2):

[0047] Total friction torque of roller bearings (1)

[0048] Frictional heat dissipation of roller bearings (2)

[0049] Among them, M rr is the rolling friction torque, M sl is the sliding friction torque, M seal is the friction torque of the seal, M drag is the friction torque caused by drag loss, eddy current and splash, and n is the spindle speed.

[0050] According to the above formulas (1) and (2), during the operation of the traction machine main shaft module, the heat consumption of the roller bearing on the traction machine main shaft is proportional to the rotation speed of the traction machine main shaft and the friction torque of the roller bearing.

[0051] Secondly, the friction torque of the roller bearing is correlated with the radial force and axial force acting on the bearing. Specifically, the calculation formulas for the friction torques of the roller bearing are shown in the following formulas (3) to (6):

[0052] (3)

[0053] (4)

[0054] (5)

[0055] (6)

[0056] Among them, D pw F is the pitch diameter of the roller assembly,r is the radial load on the bearing, v is the kinematic viscosity of the lubricant, n is the speed, R1, S1, S2 are the geometric constants of the friction torque, F a is the axial load, μ sl is the sliding friction coefficient, K s1 , β and K s2 are constants depending on the bearing type and seal type, respectively, d s is the diameter of the bearing shoulder, V M is the drag loss variable, K roll is a constant used for roller bearings and B is the width of the bearing inner ring.

[0057] According to the above formulas (3) to (6), the radial force F acting on the bearing by the main shaft is r and axial force F a It has a certain influence on the friction torque of the bearing. Therefore, when the main shaft of the traction machine is deformed, the radial force and axial force acting on the bearing will also change, causing the friction torque of the bearing to change, thereby changing the friction heat power consumption of the bearing.

[0058] Furthermore, heat flux is correlated with friction heat. Specifically, after the bearing generates heat due to friction torque, the heat flux Hc on its surface is calculated as shown in the following formula (7):

[0059] (7)

[0060] Among them, k is the thermal conductivity of the material, S is the area between two points perpendicular to the bearing in the direction of heat flow, d is the distance between two points perpendicular to the bearing, T1 is the center temperature of the contact area between the roller and the raceway, and T2 is the temperature at the axial center position of the outer surface.

[0061] Therefore, by monitoring the changes in the heat flux of the bearing, the friction torque of the bearing can be analyzed, and then the deformation of the main shaft can be inferred.

[0062] [Example 2]

[0063] Figure 2 It is a flow chart of a method for predicting the life of a main shaft module of a traction machine based on a deep convolutional neural network according to Example 2 of the present invention.

[0064] like Figure 2 As shown, in Embodiment 2 of the present invention, the life prediction method may include the following steps:

[0065] Step 1: Construction and verification of the digital model of the traction machine main shaft module

[0066] The structural parameters, operating parameters (such as load range, starting speed, operating speed, etc.) and environmental parameters (such as temperature and humidity of the space where the main shaft module is located) of the traction machine main shaft module are obtained, and a digital model of the traction machine main shaft module is generated according to the above parameters. The time variation curve of the maximum heat flux of the main shaft bearing of the traction machine main shaft module under normal operating conditions is analyzed through the digital model of the traction machine main shaft module. The analysis results are calibrated with the maximum heat flux variation curve of the main shaft bearing under the actual normal operating conditions of the traction machine main shaft module in the database, and the constructed digital model is corrected to improve the accuracy of its analysis.

[0067] Step 2: Obtain the maximum heat flux of the main shaft bearing of the traction machine main shaft module under different conditions and the corresponding time-frequency chromatogram

[0068] By setting the structural parameters of the traction machine main shaft module, the maximum heat flux time variation curve of the bearing on the traction machine main shaft is analyzed for each rotation of the traction machine main shaft, and the maximum heat flux conversion curve of the traction machine main shaft bearing in the time interval of one rotation of the traction machine main shaft module is extracted in two cases: the first case is that the main shaft is normal and the bearing is worn; the second case is that the main shaft is deformed and worn, and the bearing is normal. At the same time, the maximum heat flux time variation curve of the main shaft bearing obtained in each state is converted into a corresponding time-frequency chromatogram by wavelet transform, wherein the calculation formula of the wavelet transform is shown in the following formula (8):

[0069] (8)

[0070] Among them, f(t) represents the change of the maximum heat flux of the bearing of the traction machine main shaft with time for each rotation of the main shaft, a is the scale factor, τ is the translation factor, is the wavelet mother function.

[0071] Through the wavelet transform calculation of the above formula (8), the time-frequency chromatogram of the maximum heat flux of the bearing of the traction machine main shaft under different states is obtained.

[0072] Step 3: Fatigue life analysis of the traction machine main shaft module under different deformation and wear conditions

[0073] According to the constructed digital model of the traction machine main shaft module, the fatigue life of the traction machine main shaft bearing under different deformation and wear conditions of the main shaft is analyzed, and the fatigue life value of the main shaft bearing obtained by the analysis corresponds to the maximum heat flux time-frequency chromatogram under the main shaft state, and each fatigue life value is set as the corresponding label of each time-frequency chromatogram, thereby completing the collection of training data and test data for the life prediction (analysis) model of the traction machine main shaft module. Among them, according to the constructed digital model of the traction machine main shaft module, the ANSYS Workbench software is used to analyze the fatigue life of the corresponding bearing under different deformation and wear conditions of the main shaft, and obtain the corresponding fatigue life value of the main shaft bearing.

[0074] In an exemplary embodiment, the training data and the test data are obtained by the following method: First, the heat flux data under the following different conditions are obtained: 1. When the main shaft of the traction machine is normal and the main shaft bearing is in a normal state, the high-speed elevator is fully loaded at different starting accelerations (for example, 10 observation values ​​are calibrated between 0.5m / s² and 1.5m / s²), and the maximum heat flux value of the main shaft bearing during one rotation; 2. When the main shaft of the traction machine is normal and the main shaft bearing is in a worn state, the high-speed elevator is fully loaded at different starting accelerations (for example, 10 observation values ​​are calibrated between 0.5m / s² and 1.5m / s²). 10 observation values ​​are calibrated), the maximum heat flux value of the main shaft bearing after one rotation; 3. Different deformation and wear states of the traction machine main shaft, the main shaft is in normal state, and the high-speed elevator is fully loaded at different starting accelerations (for example, 10 observation values ​​are calibrated between 0.5m / s² and 1.5m / s²), the maximum heat flux value of the main shaft bearing after one rotation, where the different deformation and wear states of the traction machine main shaft include three deformation and wear degrees calibrated as mild, moderate and severe, which are calibrated according to the maximum equivalent stress value and maximum deformation displacement value obtained when the main shaft is loaded. Finally, the obtained heat flux data and the corresponding fatigue life value of the main shaft bearing are evenly divided into two parts, one part is used as training data, and the other part is used as test data, where the fatigue life value of the main shaft bearing is used as the label value corresponding to the heat flux data in the life prediction analysis model.

[0075] Step 4: Analysis and prediction of the life prediction model of the traction machine main shaft module based on convolutional neural network

[0076] In this embodiment, the VGG-16 network structure in the convolutional neural network is used to construct a life prediction (analysis) model for the main shaft module of the traction machine. In the life prediction (analysis) model for the main shaft module of the traction machine, an input layer, an output layer, a fully connected layer and m convolutional layers (m is an integer greater than 1) are set respectively, and a pooling layer is set for every 2 convolutional layers. This step 4 specifically includes the following steps:

[0077] (1) Data processing. The time-frequency chromatograms of the heat flux of the main shaft bearing obtained by analyzing the digital model of the traction machine main shaft module are edited into images with the same pixel size of H1×W1×G1, where H1 is the length of the image, W1 is the width of the image, and G1 is the number of channels of the image; and the corresponding labels are assigned to each time-frequency chromatogram according to the fatigue life value of the main shaft bearing corresponding to the time-frequency chromatogram, and the obtained data set is divided into two categories: training set and test set.

[0078] Among them, regarding the label setting of the time-frequency chromatogram, in some embodiments, the label is set to the life value of the spindle module in three stages, namely, the initial stage (number of uses L≥10 6 times), mid-term (use times 10 3 ≤L<10 6 times), terminal stage (number of times used L<10 3 Second-rate).

[0079] (2) Processing of the input layer. According to the pixel size of the input time-frequency spectrum, these pixels are directly brought into the neural unit of the input layer. If the pixel data at the i-th row and j-th column of the read-in image is x ij , then the output data a of the neural unit in the i-th row and j-th column of the input layer ij =x ij .

[0080] (3) Processing of the convolution layer. Set the size of the convolution kernel to F H ×F W ×F G (Among them, F H is the length of the convolution kernel, F W is the width of the convolution kernel, F G is the number of channels of the convolution kernel), in any convolution layer l, the pixel feature value of the output time-frequency chromatogram The calculation formula is shown in the following formula (9):

[0081] (9)

[0082] in, is the jth output of the lth layer, M j is the jth convolution region in the l-1 layer, i is the number of convolution calculations, represents the convolution calculation, is the j-th output of layer l-1, Represented as the corresponding weight (convolution kernel) in layer l, is the bias in layer l, and f(x) represents the ReLU nonlinear activation function.

[0083] At the same time, the size of the feature map output by the convolution layer is H2×W2×G2, where H2 is the length of the image, W2 is the width of the image, and G2 is the number of channels of the image. The sizes of H2 and W2 are calculated by the following formulas (10) and (11), respectively:

[0084] (10)

[0085] (11)

[0086] Among them, H1 is the length of the picture, W1 is the width of the picture, and F H 、F w They represent the length and width of the convolution kernel respectively, s represents the step size of the sliding window, and P represents the boundary padding value.

[0087] Through step (3), the time-frequency chromatogram of size H1×W1×G1 is input into the l-th convolutional layer, and the time-frequency chromatogram of size F H ×F W ×F G The convolution kernel is moved with a step size of s to extract the jth eigenvalue H of the chromatogram pixel j l , and output to a feature map of size H2×W2×G2 to complete a convolution calculation, where the number of channels G1, G2, F G The value of is determined according to the convolutional neural network structure model adopted.

[0088] (4) Pooling layer processing. In order to reduce the pixel data dimension of the time-frequency chromatogram after multiple convolution calculations and extract key features, the maximum pooling method shown in the following formula (12) is used to extract relevant data from the feature map:

[0089] (12)

[0090] Among them, G (x ij ) represents the result of a pooling operation, x ij represents a feature value of the i-th row and j-th column in the feature map pooling area, C y represents the yth pooling area, L H、 L W Indicates the size of the pooling area.

[0091] (5) Processing of the fully connected layer. The pixel feature values ​​of the time-frequency chromatogram extracted by the previous convolution and pooling layers are rearranged into a column, and the calculation formula is shown in the following formula (13):

[0092] (13)

[0093] in, represents the jth output of the lth layer, n represents the number of input neurons, represents the i-th output of layer l-1, represents the corresponding weight in layer l, represents the bias in layer l, and σ(x) represents the nonlinear activation function.

[0094] (6) Processing of the output layer. For the multi-classification problem of life prediction of the main shaft module of the traction machine based on the main shaft bearing, the Softmax function is used for normalization. The output channels correspond to the life of each stage one by one. Finally, the life prediction range corresponding to the channel with the maximum value is used as the network output to predict the life of the main shaft module of the traction machine. The calculation formula is shown in the following formula (14):

[0095] (14)

[0096] Among them, z i Represents the input data corresponding to the i-th data, and n represents the number of all input data.

[0097] Step 5: Evaluation of prediction results related to the prediction model

[0098] In order to effectively analyze the constructed life prediction model, the trained model is evaluated by the constructed test set, and accuracy, precision, recall and F1-score are used as evaluation criteria. The calculation formulas of each indicator are shown in the following formulas (15) to (18):

[0099] (15)

[0100] (16)

[0101] (17)

[0102] (18)

[0103] Among them, TN represents the number of negative samples predicted as negative, TP represents the number of positive samples predicted as positive, FN represents the number of positive samples predicted as negative, and FP represents the number of negative samples predicted as positive.

[0104] Step 6: Prediction of the life of the traction machine main shaft module

[0105] The maximum heat flux change of the main shaft bearing in the monitored traction machine main shaft module during one rotation is obtained through a heat flux sensor, and a heat flux change curve is generated. The corresponding time-frequency chromatogram is generated through wavelet transform. Finally, the chromatogram is input into the aforementioned traction machine main shaft module life prediction (analysis) model to predict the life of the traction machine main shaft module and obtain the predicted life data of the traction machine main shaft module.

[0106] By adopting the above-mentioned life prediction method of Example 2 of the present invention, before the traction machine main shaft module fails, based on the mutual connection relationship and the action relationship between the main shaft bearing and the main shaft, when the main shaft is not damaged, the deformation and wear of the traction machine main shaft is analyzed by analyzing the maximum heat flux of the main shaft bearing, and a life prediction model of the traction machine main shaft module is constructed by applying a deep convolutional neural network algorithm. At the same time, by constructing a digital model of the traction machine main shaft module based on the actual operation of the elevator, relevant training and test data are obtained to improve the generalization performance of the prediction model, thereby realizing accurate prediction of the life of the elevator traction machine main shaft module in different environments and types, and ensuring the safe operation of the elevator.

[0107] [Example 3]

[0108] Figure 3 It is a flow chart of a method for predicting the life of a main shaft module of a traction machine based on a deep convolutional neural network according to Example 3 of the present invention.

[0109] like Figure 3 As shown, in Embodiment 3 of the present invention, the life prediction method may include the following steps:

[0110] Step 1: Digital model construction

[0111] Obtain the traction machine main shaft module for which life prediction is required, obtain the structural parameters and operating parameters of the traction machine main shaft module in the initial operation stage of the elevator, and quickly generate a digital model of the corresponding traction machine main shaft module through the obtained relevant parameters.

[0112] Step 2: Data acquisition and processing

[0113] According to the generated digital model of the traction machine main shaft module and the above formulas (1) to (7), the maximum heat flux per rotation interval of the traction machine main shaft bearing is analyzed for two cases (the first case is that the main shaft is normal and the bearing is worn; the second case is that the main shaft is deformed and worn and the bearing is normal). According to formula (8), the heat flux conversion curve of the main shaft bearing is converted into the corresponding time-frequency chromatogram by wavelet transform. Through the fatigue analysis of the main shaft bearing, the fatigue life of the main shaft module of the traction machine main shaft under different deformation and wear conditions is predicted. According to the determined initial (number of times used L ≥ 10 6 times), mid-term (use times 103 ≤L<10 6 times), terminal stage (number of times used L<10 3 times), the life range of three stages, the spindle module life values ​​obtained by analysis are classified, and set to labels and corresponding time-frequency chromatograms to form a data set.

[0114] Step 3: Predictive Analysis Model Construction

[0115] The data set obtained in step 2 is divided into a training set and a test set. At the same time, the VGG16 network structure in the convolutional neural network is used to build a spindle module life prediction and analysis model. Five convolutional layers and three fully connected layers are set in the network structure. Each convolutional layer contains 2 to 3 convolution operation layers and a pooling layer.

[0116] Step 4: Prediction model training

[0117] In the model training stage, the time-frequency chromatogram in the training set is input into the life prediction model of the main shaft module of the traction machine, and the image size is set to 224×224×3 in the input layer. In each convolution layer, 64 convolution kernels of 3×3×64, 128 convolution kernels of 3×3×64, 256 convolution kernels of 3×3×128, 512 convolution kernels of 3×3×256, and 512 convolution kernels of 3×3×512 are set for convolution respectively, the boundary filling value P is set to 1, and the window sliding step s is set to 1. According to formulas (9) to (12), the image data is processed according to the processing flow of "convolution-->ReLU-->convolution-->ReLU-->pooling", and the corresponding feature map is output. In the fully connected layer, according to formula (13), through multiple fully connected operations, the final feature map is expanded into a one-dimensional vector, and finally 1000 neurons are output. Finally, in the output layer, the calculation results of 1000 neurons are used according to formula (14) to output 1000 predicted probability values ​​of the life of the traction machine main shaft module, and they are classified into one of the set life stages.

[0118] Step 5: Prediction model testing and evaluation

[0119] In the model testing phase, based on the relevant data of the test set, the prediction results of the traction machine main shaft module life prediction model are evaluated according to the evaluation indicators given in formulas (15) to (18), and the relevant parameters are adjusted according to the evaluation results.

[0120] Step 6: Prediction of the life of the traction machine main shaft module

[0121] The maximum heat flux change of the main shaft bearing in the monitored traction machine main shaft module during one rotation is obtained through the heat flux sensor, and a heat flux change curve is generated. The corresponding time-frequency chromatogram is generated through wavelet transform. Finally, the chromatogram is input into the aforementioned traction machine main shaft module life prediction model to predict the life of the traction machine main shaft module and obtain the predicted life data of the traction machine main shaft module.

[0122] The above-mentioned life prediction method of Example 3 of the present invention is adopted. According to the influence of the deformation of the main shaft in the traction machine main shaft module on the operation of its bearing, a digital model of the traction machine main shaft module is constructed, the interaction relationship between the main shaft bearing and the traction machine main shaft is analyzed, and a deep convolutional neural network algorithm is applied to analyze and identify the maximum heat flux of the bearing under different operating conditions of the traction machine main shaft module, so as to determine the intrinsic connection between the operating state of the main shaft bearing and the deformation state of the traction machine main shaft, and realize the effective prediction of the operating state and life of the traction machine main shaft module through the fluctuation of the maximum heat flux value of the main shaft bearing.

[0123] [Example 4]

[0124] Figure 4 : is an architecture diagram of a traction machine main shaft module life prediction system based on a deep convolutional neural network according to Embodiment 4 of the present invention. Figure 4 As shown, the life prediction system includes:

[0125] The heat flux data acquisition module 410 is used to acquire the heat flux data of the main shaft bearing in the traction machine main shaft module rotating a predetermined number of times. In this embodiment, the heat flux data acquisition module acquires the maximum heat flux data of the main shaft bearing in the traction machine main shaft module rotating one circle. The heat flux data is acquired by a heat flux sensor.

[0126] The time-frequency chromatogram generation module 420 is used to generate a heat flux variation curve according to the heat flux data, and to generate a time-frequency chromatogram according to the heat flux variation curve. In this embodiment, the heat flux variation curve is converted into a corresponding time-frequency chromatogram by wavelet transformation.

[0127] The life prediction module 430 is used to input the time-frequency chromatogram into the life prediction model of the traction machine main shaft module to obtain the predicted life data of the traction machine main shaft module.

[0128] The model acquisition module 440 is used to obtain the life prediction model of the hoisting machine main shaft module in the following manner: according to the digital model of the hoisting machine main shaft module, the maximum heat flux transformation curve of the hoisting machine main shaft rotating one circle under different conditions, the time-frequency chromatogram corresponding to the maximum heat flux transformation curve, and the fatigue life value of the main shaft bearing are obtained as a test set and a training set; according to the test set and the training set, the life prediction model of the hoisting machine main shaft module constructed based on the convolutional neural network is trained and evaluated, and the parameters of the life prediction model of the hoisting machine main shaft module are adjusted according to the evaluation results.

[0129] In this embodiment, the model acquisition module acquires the first heat flux transformation curve of the hoisting machine main shaft rotating the predetermined number of times under different states, the time-frequency chromatogram corresponding to the first heat flux transformation curve, and the fatigue life value of the main shaft bearing according to the digital model of the hoisting machine main shaft module, including: acquiring the first heat flux transformation curve of the hoisting machine main shaft rotating the predetermined number of times when the main shaft is normal and the bearing is worn according to the digital model of the hoisting machine main shaft module, the time-frequency chromatogram corresponding to the first heat flux transformation curve, and the fatigue life value of the main shaft bearing; and acquiring the first heat flux transformation curve of the hoisting machine main shaft rotating the predetermined number of times when the main shaft is deformed and worn and the bearing is normal according to the digital model of the hoisting machine main shaft module, the time-frequency chromatogram corresponding to the first heat flux transformation curve, and the fatigue life value of the main shaft bearing.

[0130] Among them, the model acquisition module generates a digital model of the traction machine main shaft module according to the structural parameters, operating parameters, and environmental parameters of the traction machine main shaft module, the operating parameters include but are not limited to: load range, starting speed, operating speed, etc., and the environmental parameters include but are not limited to: temperature and humidity of the space where the main shaft module is located, etc. Further, the model acquisition module modifies the digital model of the traction machine main shaft module in the following manner: obtaining a first change curve of the first heat flux of the main shaft bearing of the traction machine main shaft module under normal operating conditions through the digital model of the traction machine main shaft module; obtaining an actual change curve of the first heat flux of the main shaft bearing of the traction machine main shaft module under actual normal operating conditions in the database; and modifying the digital model of the traction machine main shaft module according to the first change curve and the actual change curve.

[0131] By adopting the above-mentioned life prediction system of Example 4 of the present invention, before the traction machine main shaft module fails, according to the influence of the main shaft deformation in the traction machine main shaft module on the operation of its bearings, a digital model of the traction machine main shaft module is constructed, the interaction relationship between the main shaft bearing and the traction machine main shaft is analyzed, and a deep convolutional neural network algorithm is applied to analyze and identify the heat flux of the bearing under different operating conditions of the traction machine main shaft module, so as to determine the intrinsic connection between the operating state of the main shaft bearing and the deformation state of the traction machine main shaft, and it is possible to effectively predict the operating state and life of the traction machine main shaft module through the fluctuation of the heat flux value of the main shaft bearing, so as to ensure the reliability and safety of the elevator operation.

[0132] Through the description of the above implementation modes, those skilled in the art can clearly understand that the present invention can be implemented by combining software with a hardware platform. Based on such an understanding, all or part of the contribution of the technical solution of the present invention to the background technology can be embodied in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various implementation modes of the present invention or some parts of the implementation modes.

[0133] Correspondingly, an embodiment of the present invention further provides a computer-readable storage medium, on which computer-readable instructions or programs are stored, and when the computer-readable instructions or programs are executed by a processor, the computer performs the following operations: the operations include the steps included in the life prediction method described in any of the above embodiments, which are not repeated here. The storage medium may include, for example, a CD, a hard disk, a floppy disk, a flash memory, a magnetic tape, etc.

[0134] In addition, an embodiment of the present invention further provides a computer device including a memory and a processor, wherein the memory is used to store one or more computer-readable instructions or programs, wherein the one or more computer-readable instructions or programs can implement the life prediction method described in any of the above embodiments when executed by the processor. The computer device can be, for example, a server, a desktop computer, a notebook computer, a tablet computer, etc.

[0135] The embodiment of the present invention also provides a computer program product including a computer program, wherein the computer program includes a program code for executing the life prediction method shown in the flowchart. When the computer program product is run in a computer system, the program code is used to enable the computer system to implement the life prediction method provided by the embodiment of the present disclosure.

[0136] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention is described in detail with reference to the above embodiments, a person skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention. Therefore, the protection scope of the present invention shall be subject to the claims.

Claims

1. A method for predicting the life of a traction machine main shaft module based on a deep convolutional neural network, characterized in that: The life prediction method comprises: Acquiring heat flux data of a main shaft bearing in the traction machine main shaft module rotating a predetermined number of times; generating a heat flux variation curve according to the heat flux data; generating a time-frequency chromatogram according to the heat flux variation curve; Inputting the time-frequency chromatogram into a life prediction model of a main shaft module of a traction machine to obtain predicted life data of the main shaft module of the traction machine; The life prediction model of the traction machine main shaft module is obtained by the following method: According to the digital model of the traction machine main shaft module, a first heat flux conversion curve of the traction machine main shaft rotating the predetermined number of revolutions when the main shaft is normal and the bearing is worn and when the main shaft is deformed and worn and the bearing is normal, a time-frequency chromatogram corresponding to the first heat flux conversion curve, and a fatigue life value of the main shaft bearing are obtained as a test set and a training set; The life prediction model of the traction machine main shaft module constructed based on the convolutional neural network is trained and evaluated according to the test set and the training set, and the parameters of the life prediction model of the traction machine main shaft module are adjusted according to the evaluation results.

2. The life prediction method according to claim 1, characterized in that: A digital model of the traction machine main shaft module is generated according to the structural parameters, operating parameters and environmental parameters of the traction machine main shaft module.

3. The life prediction method according to claim 2, characterized in that: Acquire a first variation curve of a first heat flux of a main shaft bearing of the traction machine main shaft module under normal operation by using a digital model of the traction machine main shaft module; Obtaining an actual variation curve of the first heat flux of the main shaft bearing of the traction machine main shaft module under actual normal operation in the database; The digital model of the traction machine main shaft module is corrected according to the first change curve and the actual change curve.

4. A traction machine main shaft module life prediction system based on deep convolutional neural network, characterized in that: The life prediction system comprises: A heat flux data acquisition module, used to acquire heat flux data of a main shaft bearing in the traction machine main shaft module rotating a predetermined number of revolutions; A time-frequency chromatogram generation module, used to generate a heat flux variation curve according to the heat flux data, and to generate a time-frequency chromatogram according to the heat flux variation curve; A life prediction module, used for inputting the time-frequency chromatogram into a life prediction model of a main shaft module of a traction machine to obtain predicted life data of the main shaft module of the traction machine; The model acquisition module is used to obtain the life prediction model of the traction machine main shaft module in the following way: According to the digital model of the traction machine main shaft module, a first heat flux conversion curve of the traction machine main shaft rotating the predetermined number of revolutions when the main shaft is normal and the bearing is worn and when the main shaft is deformed and worn and the bearing is normal, a time-frequency chromatogram corresponding to the first heat flux conversion curve, and a fatigue life value of the main shaft bearing are obtained as a test set and a training set; The life prediction model of the traction machine main shaft module constructed based on the convolutional neural network is trained and evaluated according to the test set and the training set, and the parameters of the life prediction model of the traction machine main shaft module are adjusted according to the evaluation results.

5. A computer-readable storage medium storing computer-readable instructions, characterized in that: The computer-readable instructions are executed by a processor to implement a method for predicting the life of a traction machine main shaft module based on a deep convolutional neural network as described in any one of claims 1-3.

6. A computer device comprising a memory and a processor, The memory stores computer-readable instructions, characterized in that: The processor executes the computer-readable instructions to implement the traction machine main shaft module life prediction method based on deep convolutional neural network as described in any one of claims 1-3.

7. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, it implements the method for predicting the life of a traction machine main shaft module based on a deep convolutional neural network as described in any one of claims 1-3.

Images