Engine connecting rod big end bearing life prediction method, device and equipment

Abstract

The present application discloses a method, device, and equipment for predicting the life of an engine connecting rod big end bearing, which relates to the field of engine technology. The method comprises: determining the physical model fatigue factor of the connecting rod big end bearing based on the fatigue analysis physical model of the engine connecting rod big end bearing, and mapping it to an initial category label; performing one-dimensional deep feature extraction on the collected segmented vibration signal to obtain a feature vector; initializing the classification unit and clustering unit based on the initial category label, determining the loss function between the input initial category label and the output category label, and updating the network parameters; calculating the divergence difference, assigning dynamic weights, and correcting the initial category label to obtain the output results of the target classification unit and the target clustering unit, thereby generating a life prediction result for the connecting rod big end bearing. By combining the advantages of deep learning and physical models in the above manner, the life of the engine connecting rod big end bearing is predicted, which can improve the accuracy and reliability of the prediction.

Description

Technical Field

[0001] The present application relates to the field of engine technology, and in particular to a method, device and equipment for predicting the life of an engine connecting rod big end bearing. Background Art

[0002] The connecting rod big-end bearing is a key component connecting the connecting rod big end to the crankshaft. It transmits the combustion pressure of the piston to the crankshaft, enabling power conversion and output. It also supports the rotation of the crankshaft, ensuring stability and accuracy at high speeds and reducing vibration and wear caused by unbalanced forces. Any failure of the connecting rod big-end bearing can cause serious engine damage, such as bearing damage, crankshaft bending or breakage, or even complete engine failure. Therefore, ensuring the reliable durability and lifespan of the connecting rod big-end bearing is crucial to the long-term stable operation, reliability, and safety of the engine.

[0003] At present, for the life prediction of connecting rod big end bearings, the traditional empirical formula method has limited accuracy, is difficult to adapt to the complex and changeable actual working conditions, is overly dependent on historical data, and is not accurate enough for bearing life prediction under new engines or special working conditions; the traditional finite element analysis method uses engineering software to perform mechanical analysis to predict the stress and strain distribution of connecting rod big end bearings under specific working conditions, including model establishment, boundary setting, model debugging and operation calculations. It requires a large amount of simulation resources and has a long cycle and high computational cost, making it unsuitable for real-time monitoring and rapid prediction; traditional machine learning methods are highly dependent on feature engineering, and have difficulty processing high-dimensional data and capturing complex patterns in time series, and their accuracy needs to be improved.

[0004] The above content is only used to assist in understanding the technical solution of the present invention and does not constitute an admission that the above content is prior art. Summary of the Invention

[0005] The main purpose of this application is to provide a method, device and equipment for predicting the life of an engine connecting rod big end bearing, aiming to solve the technical problem of low accuracy in predicting the life of an engine connecting rod big end bearing in the prior art.

[0006] To achieve the above objectives, the present application provides a method for predicting the life of an engine connecting rod big end bearing, the method comprising:

[0007] Based on the fatigue analysis physical model of the engine connecting rod big end bearing, the physical model fatigue factor of the connecting rod big end bearing is determined, and the physical model fatigue factor is mapped to the initial category label;

[0008] Obtain vibration signals at different times throughout the life cycle of the connecting rod big end bearing, perform one-dimensional deep feature extraction on the segmented vibration signals, and obtain the corresponding feature vectors of the vibration signals;

[0009] Initialize the classification unit and clustering unit based on the initial category label, input the feature vector into the classification unit and clustering unit, determine the loss function between the initial category label and the output category label, and update the network parameters of the classification unit and clustering unit;

[0010] Assign dynamic weights based on the divergence difference between the initial category labels and the category labels;

[0011] Based on the dynamic weight, the initial category label is modified, the loss function and dynamic weight are re-determined, and the iteration is continued until the termination condition is met to obtain the target classification unit and target clustering unit with the target initial category label;

[0012] Based on the output results of the target classification unit and the target clustering unit, the life prediction result of the connecting rod big end bearing is determined.

[0013] In one embodiment, based on the fatigue analysis physical model of the engine connecting rod big end bearing, the step of determining the fatigue factor of the physical model of the connecting rod big end bearing further includes:

[0014] Obtain fatigue influencing parameters of the connecting rod big end bearing, the fatigue influencing parameters at least including local maximum stress, nominal stress, average stress, ultimate tensile strength of the bearing material, fatigue strength of a standard test piece, fatigue strength of an actual sample, surface fatigue strength of a smooth standard test piece, surface fatigue strength of an actual sample, symmetric fatigue limit of the bearing material, and sensitivity coefficient of the bearing material to stress cycle asymmetry;

[0015] Based on the first correspondence between fatigue influencing parameters and fatigue factors of the physical model, a fatigue analysis physical model is constructed.

[0016] In one embodiment, the step of performing one-dimensional deep feature extraction on the segmented vibration signal to obtain a feature vector corresponding to the vibration signal includes:

[0017] The segmented vibration signal is subjected to one-dimensional deep feature extraction through a one-dimensional deep self-encoding feature unit to obtain a feature vector corresponding to the vibration signal. The one-dimensional deep self-encoding feature unit includes an encoding network and a decoding network. The encoding network includes a convolution layer and a pooling layer, and the decoding network includes an upsampling layer and a convolution layer.

[0018] In one embodiment, the category labels include a first category label and a second category label, and the steps of inputting the feature vector into the classification unit and the clustering unit, determining a loss function between the initial category label and the output category label, and updating the network parameters of the classification unit and the clustering unit include:

[0019] Input the feature vector into the classification unit and the clustering unit to obtain the first category label output by the classification unit and the second category label output by the clustering unit respectively;

[0020] Calculate the classification loss based on the initial category label and the first category label;

[0021] Calculate the clustering loss based on the initial category label and the second category label;

[0022] Determine the loss function based on classification loss and clustering loss;

[0023] Based on the gradient information of the loss function, the network parameters of the classification unit and clustering unit are updated.

[0024] In one embodiment, the step of determining the loss function based on the classification loss and the clustering loss includes:

[0025] Obtain the reconstruction loss of vibration signals during one-dimensional deep feature extraction;

[0026] Obtain the second correspondence between classification loss, clustering loss, reconstruction loss, balance coefficient and loss function;

[0027] Based on the classification loss, clustering loss, reconstruction loss, balance coefficient and the second correspondence, a loss function between the initial category label and the category label is determined.

[0028] In one embodiment, before the step of assigning a dynamic weight based on the divergence difference between the initial category label and the category label, the step further includes:

[0029] Based on the category labels and the initial category labels, determine the label mean;

[0030] Obtain the third correspondence between the initial category label, label mean and divergence difference;

[0031] Based on the initial category labels, the label means, and the third correspondence, a divergence difference between the initial category labels and the category labels is determined.

[0032] In one embodiment, the step of assigning a dynamic weight based on the divergence difference between the initial category label and the category label includes:

[0033] Obtaining a fourth correspondence between the divergence difference, the adjustment parameter, and the dynamic weight;

[0034] A dynamic weight is determined based on a divergence difference between the initial category label and the category label, an adjustment parameter, and a fourth corresponding relationship.

[0035] In one embodiment, the step of revising the initial category labels based on the dynamic weights includes:

[0036] Obtaining a fifth correspondence between the initial category label, the category label, the category label weight, the dynamic weight, the transformation coefficient, and the revised initial category label;

[0037] A revised initial category label is determined based on the initial category label, the category label, the weight of the category label, the dynamic weight, the transformation coefficient, and the fifth corresponding relationship.

[0038] In addition, to achieve the above-mentioned purpose, the present application also proposes a device for predicting the life of an engine connecting rod big end bearing, the device comprising:

[0039] An analysis module is used to determine a fatigue factor of the connecting rod big end bearing based on a fatigue analysis physical model of the engine connecting rod big end bearing, and map the fatigue factor of the physical model to an initial category label;

[0040] The correction module is used to obtain the vibration signals of the connecting rod big end bearing at different times during its entire life cycle, perform one-dimensional deep feature extraction on the segmented vibration signals, and obtain the feature vector corresponding to the vibration signal;

[0041] The correction module is further used to initialize the classification unit and the clustering unit based on the initial category label, input the feature vector into the classification unit and the clustering unit, determine the loss function between the initial category label and the output category label, and update the network parameters of the classification unit and the clustering unit;

[0042] The correction module is also used to assign dynamic weights based on the divergence difference between the initial class labels and the class labels;

[0043] The correction module is also used to correct the initial category labels based on the dynamic weights, redefine the loss function and dynamic weights, and iterate continuously until the termination conditions are met to obtain the target classification unit and target clustering unit with the target initial category labels;

[0044] The prediction module is used to determine the life prediction result of the connecting rod big end bearing based on the output results of the target classification unit and the target clustering unit.

[0045] In addition, to achieve the above-mentioned purpose, the present application also proposes an engine connecting rod big end bearing life prediction device, which includes: a memory, a processor, and a computer program stored in the memory and runnable on the processor, and the computer program is configured to implement the steps of the engine connecting rod big end bearing life prediction method as described above.

[0046] In addition, to achieve the above-mentioned purpose, the present invention also proposes a storage medium, which is a computer-readable storage medium and stores a computer program. When the computer program is executed by a processor, the steps of the engine connecting rod big end bearing life prediction method as described above are implemented.

[0047] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the steps of the engine connecting rod big end bearing life prediction method as described above.

[0048] The present application provides a life prediction method for an engine connecting rod big end bearing. Based on the fatigue analysis physical model of the engine connecting rod big end bearing, the physical model fatigue factor of the connecting rod big end bearing is determined, and the physical model fatigue factor is mapped to an initial category label; the vibration signals of the connecting rod big end bearing at different times in the entire life cycle are obtained, and one-dimensional deep feature extraction is performed on the segmented vibration signal to obtain a feature vector corresponding to the vibration signal; the classification unit and the clustering unit are initialized based on the initial category label, the feature vector is input into the classification unit and the clustering unit, the loss function between the initial category label and the output category label is determined, and the network parameters of the classification unit and the clustering unit are updated; dynamic weights are assigned based on the divergence difference between the initial category label and the category label; based on the dynamic weights, the initial category label is corrected, the loss function and the dynamic weights are re-determined, and the iteration is continued until the termination condition is met to obtain a target classification unit and a target clustering unit with the target initial category label; based on the output results of the target classification unit and the target clustering unit, the life prediction result of the connecting rod big end bearing is determined. Based on the physical analysis model, this application assigns an initial life label to the collected vibration signal of the connecting rod big end bearing, then extracts the deep features of the vibration signal, and finally based on the KL divergence, continuously and adaptively corrects the life label during the model training process, and finally obtains a more stable and accurate life prediction result. It combines the advantages of deep learning and physical models, improves the accuracy and reliability of the prediction, and through multi-dimensional analysis, considers the actual operating conditions and material properties of the equipment, can adapt to the life prediction needs under different working conditions and usage conditions, reduces dependence on manual experience, has good versatility and adaptability, and solves the technical problem of low accuracy in life prediction of engine connecting rod big end bearings. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0050] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0051] Figure 1 This is a flow chart of Example 1 of the engine connecting rod big end bearing life prediction method of the present application;

[0052] Figure 2 A schematic diagram of the coding network structure of the engine connecting rod big end bearing life prediction method provided in Example 1 of the present application;

[0053] Figure 3 A schematic diagram of the decoding network structure of the engine connecting rod big end bearing life prediction method provided in Example 1 of the present application;

[0054] Figure 4 This is a schematic diagram of the overall structure of the engine connecting rod big end bearing life prediction method provided in Example 1 of the present application;

[0055] Figure 5 A schematic diagram of a simplified flow chart of a method for predicting the life of an engine connecting rod big end bearing provided in Example 1 of the present application;

[0056] Figure 6 This is a schematic diagram of the module structure of the engine connecting rod big end bearing life prediction device according to an embodiment of the present application;

[0057] Figure 7 Schematic diagram of the equipment structure of the hardware operating environment involved in the engine connecting rod big end bearing life prediction method in the embodiment of the present application.

[0058] The realization of the objectives, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0059] It should be understood that the specific embodiments described herein are merely used to explain the technical solutions of the present application and are not intended to limit the present application.

[0060] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0061] The main solution of the embodiment of the present application is: based on the fatigue analysis physical model of the engine connecting rod big end bearing, determine the physical model fatigue factor of the connecting rod big end bearing, and map the physical model fatigue factor to an initial category label; obtain the vibration signals of the connecting rod big end bearing at different times in the entire life cycle, perform one-dimensional deep feature extraction on the segmented vibration signal, and obtain the feature vector corresponding to the vibration signal; initialize the classification unit and clustering unit based on the initial category label, input the feature vector into the classification unit and clustering unit, determine the loss function between the initial category label and the output category label, and update the network parameters of the classification unit and clustering unit; assign dynamic weights based on the divergence difference between the initial category label and the category label; based on the dynamic weight, correct the initial category label, re-determine the loss function and dynamic weight, and iterate continuously until the termination condition is met, and obtain the target classification unit and target clustering unit with the target initial category label; based on the output results of the target classification unit and the target clustering unit, determine the life prediction result of the connecting rod big end bearing.

[0062] This application provides a solution. According to the physical analysis model, an initial life label is assigned to the collected vibration signal of the connecting rod big end bearing, and then the deep features of the vibration signal are extracted. Finally, based on the KL divergence, the life label is continuously and adaptively corrected during the model training process, and finally a more stable and accurate life prediction result is obtained. It combines the advantages of deep learning and physical models, improves the accuracy and reliability of the prediction, and through multi-dimensional analysis, considers the actual operating conditions and material properties of the equipment, can adapt to the life prediction needs under different working conditions and usage conditions, reduces dependence on manual experience, has good versatility and adaptability, and solves the technical problem of low accuracy in life prediction of engine connecting rod big end bearings.

[0063] It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication, and program execution capabilities, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the aforementioned functions, such as an engine connecting rod big end bearing life prediction device, and this embodiment does not specifically limit this. The following uses an engine connecting rod big end bearing life prediction device as an example to illustrate this embodiment and the following embodiments.

[0064] The present invention provides a method for predicting the life of an engine connecting rod big end bearing, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the engine connecting rod big end bearing life prediction method of the present application.

[0065] In this embodiment, the engine connecting rod big end bearing life prediction method includes steps S10 to S60:

[0066] Step S10, determining a physical model fatigue factor of the connecting rod big end bearing based on a fatigue analysis physical model of the engine connecting rod big end bearing, and mapping the physical model fatigue factor to an initial category label;

[0067] It should be noted that the fatigue analysis physical model of the engine connecting rod big end bearing is a calculation model used to perform fatigue analysis on the connecting rod big end bearing and evaluate the fatigue safety factor. In this embodiment, the fatigue factor of the physical model of the connecting rod big end bearing, that is, the fatigue safety factor of the connecting rod big end bearing, serves as an evaluation indicator for the life of the connecting rod big end bearing.

[0068] In a feasible implementation, the step of constructing the fatigue analysis physical model includes: obtaining fatigue influencing parameters of the connecting rod big end bearing; and constructing the fatigue analysis physical model based on a first correspondence between the fatigue influencing parameters and the fatigue factors of the physical model.

[0069] It should be noted that fatigue influencing parameters are parameters that have an impact on fatigue analysis. Based on the preset finite element model / material properties, the fatigue influencing parameters of the connecting rod big end bearing are obtained. The fatigue influencing parameters include at least the local maximum stress, nominal stress, average stress, the tensile strength of the bearing material, the fatigue strength of the standard test piece, the fatigue strength of the actual sample, the surface fatigue strength of the smooth standard test piece, the surface fatigue strength of the actual sample, the symmetrical fatigue limit of the bearing material, and the sensitivity coefficient of the bearing material to stress cycle asymmetry. During engine operation, the connecting rod big end bearing is mainly subjected to the cyclical alternating loads generated by the gas combustion pressure and reciprocating inertia force, which is prone to fatigue damage. Therefore, the fatigue influencing parameters should take into account multiple factors such as workload, operating conditions, lubrication method, lubrication status, material properties, and geometric structure to provide comprehensive input data for the life prediction and performance evaluation of the connecting rod big end bearing.

[0070] It should be noted that the first correspondence between fatigue influencing parameters and physical model fatigue factors is as follows:

[0071]

[0072] Where SF represents the physical model fatigue factor, σ max represents the local maximum stress, σ1 represents the nominal stress, σ S1 Represents the surface fatigue strength of the real sample, σ S2 It represents the surface fatigue strength of the smooth standard test piece, S1 represents the fatigue strength of the real sample, and S2 represents the fatigue strength of the standard test piece. Indicates the sensitivity coefficient of the bearing material to the asymmetric stress cycle, σ α represents the stress amplitude, σ m represents the mean stress, σ -1This embodiment constructs a fatigue analysis physical model based on the first corresponding relationship.

[0073] In addition, it should be noted that the physical model fatigue factor SF is used to characterize the fatigue influencing parameters of the connecting rod big end bearing and the influence of the bearing material properties on the bearing life. It represents the theoretical stress concentration coefficient, β represents the surface strengthening coefficient, and ε σ represents the size influence coefficient, Indicates the sensitivity coefficient of the bearing material to the asymmetric stress cycle, σ α represents the stress amplitude, σ m represents the mean stress, σ -1 Represents the symmetrical fatigue limit of the bearing material.

[0074] It can be understood that the physical model fatigue factor σ is used to characterize the fatigue influencing parameters of the connecting rod big end bearing and the influence of the bearing material properties on the bearing life. Also known as the theoretical stress concentration factor, it is used to characterize the local stress concentration phenomenon caused by notches, chamfers, fillets, etc. in the structural design of parts, and its influence on the fatigue life of bearings. max is the local maximum stress, σ1 is the nominal stress. Also known as the surface strengthening coefficient, it is used to characterize the influence of the surface roughness, surface treatment state, and surface treatment method on the fatigue strength of the connecting rod big end bearing. Generally, the worse the bearing surface condition, the lower the fatigue strength. S1 is the surface fatigue strength of the real sample, σ S2 The surface fatigue strength of the smooth standard test piece. Due to the difference between the actual part size and the standard test piece size, size correction is required, so the ratio is set. Also known as the size influence coefficient, it is used to characterize the influence of the size structure of the bearing on the fatigue strength. S1 is the fatigue strength of the real sample, and S2 is the fatigue strength of the standard test piece. The sensitivity coefficient of the bearing material to the asymmetry of the stress cycle Depends on the bearing material properties and is provided by the supplier. Also known as stress amplitude, it is used to characterize the characteristics of the stress amplitude of the connecting rod big end bearing. The stress amplitude can be obtained according to the preset connecting rod finite element model, or it can be estimated according to the fitting formula. There are multiple estimation models for the stress amplitude of the connecting rod big end bearing, which can be flexibly adjusted according to actual conditions. There is no specific limitation on this. In this embodiment, the fitting formula estimation method is adopted, and the estimation formula is: σ b is the ultimate tensile strength of the bearing material, σ m is the mean stress, σ -1 is the symmetrical fatigue limit of the bearing material. The tensile strength of the bearing material σ bRelated to the material properties, provided by the supplier, the mean stress σ m It can be obtained based on the preset connecting rod finite element model. The symmetrical fatigue limit σ of the bearing material -1 It can be obtained by experimental measurement or estimation based on fitting formula. For example, it can be calculated based on the static strength σ of the material. If it is steel material and the static strength σ is less than 1800MPa, then σ -1 =0.5σ, if it is magnesium, copper or nickel alloy material, then σ -1 =0.35σ, if it is aluminum alloy material, then σ -1 =1.49σ 0.63 .

[0075] Since fatigue analysis is a complex process influenced by multiple factors, in practical applications, fatigue influencing parameters can be flexibly adjusted according to actual needs. The formula for the fatigue factor of the physical model of the connecting rod big end bearing may vary depending on the specific situation, and a more complex model may be required for accurate prediction. It should be flexibly adjusted according to actual conditions to ensure the reliability and performance optimization of the connecting rod big end bearing design.

[0076] It should be understood that the physical model fatigue factor is mapped to the lifespan label defined in the actual situation to obtain the initial category label. The lifespan label can be set according to the actual situation. For example, based on the year as the basic unit, less than 1 year is classified into one category, more than 1 year but less than 5 years is classified into another category, 5-10 years is classified into another category, and more than 10 years is classified into another category. There is no specific limitation on this. The initial category label at this time is a preliminary label and needs to be further revised later.

[0077] Step S20, obtaining vibration signals of the connecting rod big end bearing at different times during the entire life cycle, performing one-dimensional deep feature extraction on the segmented vibration signals, and obtaining feature vectors corresponding to the vibration signals;

[0078] It should be noted that this embodiment collects vibration signals at different times throughout the life cycle of an engine connecting rod big end bearing for use in connecting rod big end bearing life prediction. The vibration signal data can be angular velocity data or angular acceleration data collected by a sensor, and can be flexibly adjusted based on actual conditions, without specific limitation.

[0079] In addition, it should be noted that the collected vibration signals need to be segmented and processed. The length of the segments can be set according to actual needs and is not specifically limited. In this embodiment, the collected vibration signals are segmented and intercepted according to a length of 2560.

[0080] It can be understood that the feature vector is a vector composed of the features of the vibration signal. The segmented vibration signal is subjected to one-dimensional deep feature extraction to obtain the corresponding feature vector

[0081] In a feasible embodiment, the step of performing one-dimensional deep feature extraction on the segmented vibration signal to obtain a feature vector corresponding to the vibration signal includes: performing one-dimensional deep feature extraction on the segmented vibration signal through a one-dimensional deep self-encoding feature unit to obtain a feature vector corresponding to the vibration signal.

[0082] It should be noted that this embodiment performs one-dimensional deep feature extraction through a one-dimensional deep self-encoding feature unit. The one-dimensional deep self-encoding feature unit includes an encoding network (encoding module) and a decoding network (decoding module). The encoding network includes a convolution layer and a pooling layer, and the decoding network includes an upsampling layer and a convolution layer. The number of convolution layers and pooling layers in the encoding network can be set according to actual needs, and the number of upsampling layers and convolution layers in the decoding network can be set according to actual needs, and there is no specific limitation on this.

[0083] For example, refer to Figure 2 , 6 convolutional layers (Conv1~Conv6) and 4 pooling layers (Pool1~Pool4) are set in the encoding module, refer to Figure 3 , 6 convolutional layers (Conv6~Conv12) and 4 upsampling layers (Upsample1~Upsample4) are set in the decoding module.

[0084] Step S30, initializing the classification unit and the clustering unit based on the initial category label, inputting the feature vector into the classification unit and the clustering unit, determining the loss function between the initial category label and the output category label, and updating the network parameters of the classification unit and the clustering unit;

[0085] It should be noted that the classification unit and clustering unit are initialized using the initial category labels. After initialization, the feature vectors are input into the classification unit and clustering unit. The classification unit is the unit in the deep learning model responsible for mapping the learned features to specific categories. The clustering unit classifies the feature vectors into different specific categories by learning the relationship between the feature vectors. Ultimately, the classification unit and clustering unit will output the corresponding category probabilities, thereby determining the category labels corresponding to the classification unit and clustering unit. The initial category labels and the category labels of the classification unit and clustering unit are then used to calculate the loss function.

[0086] In a feasible implementation, step S30 may include steps S301 to S304:

[0087] Step S301: Input the feature vector into the classification unit and the clustering unit to obtain the first category label output by the classification unit and the second category label output by the clustering unit respectively;

[0088] It should be noted that the category label includes the first category label and the second category label. The category label output by the classification unit is used as the first category label, and the category label output by the clustering unit is used as the second category label. Figure 4 , input the feature vector into the classification unit and clustering unit respectively, and the classification unit outputs the first category label l c , the clustering unit outputs the second category label l cl .

[0089] Step S302: Calculate the classification loss based on the initial category label and the first category label, and calculate the clustering loss based on the initial category label and the second category label;

[0090] It should be noted that the classification loss is the loss L corresponding to the difference between the initial category label and the first category label output by the classification unit. c , the clustering loss is the loss L corresponding to the difference between the initial category label and the second category label output by the clustering unit cl .

[0091] Step S303: determining a loss function based on the classification loss and the clustering loss;

[0092] In a feasible implementation, step S303 may include: obtaining the reconstruction loss of the vibration signal in the one-dimensional deep feature extraction process; obtaining the second correspondence between the classification loss, clustering loss, reconstruction loss, balance coefficient and the loss function; and determining the loss function between the initial category label and the category label based on the classification loss, clustering loss, reconstruction loss, balance coefficient and the second correspondence.

[0093] It should be noted that the reconstruction loss is the loss calculated between the reconstructed signal and the original vibration signal. The original vibration signal is the segmented vibration signal input into the one-dimensional deep self-encoding feature unit, and the reconstructed signal is the reconstructed vibration signal. Figure 4 The original vibration signal can be considered as the input signal L1 of the first convolution layer Conv1 of the encoding unit, and the reconstructed signal can be considered as the output signal L2 of the last convolution layer Conv12 of the decoding unit. The loss between signals L1 and L2 is the reconstruction loss L x .

[0094] It can be understood that the second corresponding relationship between classification loss, clustering loss, reconstruction loss, balance coefficient and loss function is the calculation relationship of the loss function, which is as follows:

[0095] L=L x +λ(L c +L cl )

[0096] Where L represents the loss function, Lc represents the classification loss, L cl represents the clustering loss, L x Denotes the reconstruction loss, and λ denotes the balance coefficient. The balance coefficient λ is a weight used to balance the classification loss and clustering loss. In the early stages of training, due to the instability of the classification and clustering unit performance, λ is assigned a lower weight. However, as the classification performance of the classification and clustering units stabilizes, λ is assigned a higher weight. Substituting the classification loss, clustering loss, reconstruction loss, and balance coefficient into the second correspondence above, the loss function between the initial class label and the class label is calculated.

[0097] Step S304: Update the network parameters of the classification unit and the clustering unit based on the gradient information of the loss function.

[0098] It should be understood that after obtaining the loss function, the gradient of the loss function with respect to the network parameters of the classification unit and clustering unit is calculated and passed back to the network through the back propagation algorithm. An optimization algorithm (such as SGD or Adam) can be used to update the network parameters of the classification unit and clustering unit based on the gradient information.

[0099] Step S40, assigning a dynamic weight based on the divergence difference between the initial category label and the category label;

[0100] It should be noted that this embodiment uses KL divergence to calculate the difference between the category label and the initial category label, and allocates dynamic weights accordingly. The calculated difference is the divergence difference.

[0101] In a feasible implementation, the step of calculating the divergence difference may include: determining the label mean based on the category label and the initial category label; obtaining a third correspondence between the initial category label, the label mean and the divergence difference; and determining the divergence difference between the initial category label and the category label based on the initial category label, the label mean and the third correspondence.

[0102] It should be noted that the label mean is the mean of the category label and the initial category label, that is, the mean between the first category label, the second category label and the initial category label. The calculation relationship is as follows:

[0103]

[0104] Where, represents the label mean, l c represents the first category label, l cl represents the second category label, and l represents the initial category label.

[0105] It can be understood that the third correspondence between the initial category label, the label mean and the divergence difference is the calculation relationship of the divergence difference, which is as follows:

[0106]

[0107] Where D KL (l|l) represents the divergence difference, l i Denotes the initial category label of the i-th category, and denotes the mean label of the i-th category. Substitute the relevant data into the data and calculate the divergence difference between the category label and the initial category label, so as to assign dynamic weights according to the divergence difference.

[0108] In a feasible implementation, step S40 may include: obtaining a fourth correspondence between the divergence difference, the adjustment parameter and the dynamic weight; and determining the dynamic weight based on the divergence difference between the initial category label and the category label, the adjustment parameter and the fourth correspondence.

[0109] It should be noted that the fourth corresponding relationship between the divergence difference, the adjustment parameter and the dynamic weight is the dynamic weight calculation relationship, which is as follows:

[0110]

[0111] In the formula, β represents the dynamic weight, Denotes the divergence difference, and τ is the adjustment parameter. τ is a positive adjustment parameter that controls the rate of change of the dynamic weight. When τ is large, the dynamic weight is more sensitive to changes in the divergence difference, while when τ is small, the dynamic weight changes more slowly in response to the divergence difference. Substituting the divergence difference and the adjustment parameter into the fourth correspondence relationship above, the dynamic weight is calculated.

[0112] Step S50: Based on the dynamic weight, the initial category label is modified, the loss function and the dynamic weight are re-determined, and the iteration is continued until the termination condition is met, thereby obtaining the target classification unit and the target clustering unit with the target initial category label;

[0113] It can be understood that dynamic weights can reduce the impact of errors on label correction.

[0114] In a feasible embodiment, the step of correcting the initial category label based on the dynamic weight includes: obtaining the fifth correspondence between the initial category label, the category label, the weight of the category label, the dynamic weight, the transformation coefficient and the corrected initial category label; and determining the corrected initial category label based on the initial category label, the category label, the weight of the category label, the dynamic weight, the transformation coefficient and the fifth correspondence.

[0115] It should be noted that the fifth corresponding relationship between the initial category label, the category label, the category label weight, the dynamic weight, the transformation coefficient and the revised initial category label is the calculation relationship used to revise the initial category label, as shown below;

[0116]

[0117] In the formula, l* represents the initial category label after correction, β represents the dynamic weight, l represents the initial category label before correction, l c represents the first category label, β c represents the weight of the first category label, l cl represents the second category label, β cl Represents the weight of the second category label, w represents the transformation coefficient. Weight β c and β cl It can be set according to the actual situation and there is no specific limitation on this. In addition, when the value of the loss function is large, the weight β is usually adjusted. c and β cl , making the outputs of the classification unit and clustering unit closer to the true label. The transformation coefficient w is linearly transformed from 0 to 1 during the correction process. This does not give the new label a high weight at the beginning, which can achieve the purpose of stable correction.

[0118] Understandably, the reference Figure 4 , using the fifth corresponding relationship mentioned above, determine the corrected initial category label l*, and update the previous initial category label l. At this time, it is necessary to determine whether the termination condition is met. The termination condition can be reaching the maximum number of iterations or meeting the accuracy requirements, which is not specifically limited. If the termination condition is met, it means that the classification unit and clustering unit at this time have been trained and can be used to predict the life of the connecting rod big end bearing. If the termination condition is not met, it means that the classification unit and clustering unit at this time have not been trained. It is necessary to continue training, redefine the loss function and dynamic weight, re-correct the initial category label, and iterate continuously until the termination condition is met. The final initial category label is the target initial category label, the classification unit that is finally trained is the target classification unit, and the clustering unit that is finally trained is the target clustering unit.

[0119] It should be understood that when it is determined that the termination condition is not met, the network parameters of the classification unit and clustering unit can be fine-tuned first, and then return to the previous step to redefine the loss function and dynamic weights, and re-correct the initial category labels.

[0120] In this embodiment, dynamic weights are determined based on KL divergence, a dynamic weight allocation mechanism based on prediction reliability and importance. Labels can be dynamically modified during training, giving the model adaptive learning capabilities and improving prediction accuracy and robustness. This mechanism not only enhances the model's generalization capabilities for new or unknown situations, but also reduces reliance on large-scale labeled datasets, reducing the complexity and cost of data preparation. Furthermore, the nature of unsupervised learning further enhances the model's generalization performance, enabling it to adapt to changing industrial environments while optimizing maintenance strategies and achieving cost-effectiveness.

[0121] Step S60: Determine the life prediction result of the connecting rod big end bearing based on the output results of the target classification unit and the target clustering unit.

[0122] It can be understood that the target classification unit and the target clustering unit are used to perform the final lifespan prediction and obtain the corresponding lifespan prediction result.

[0123] It should be understood that this embodiment significantly improves the accuracy of connecting rod big-end bearing life prediction through the innovative integration of deep learning models and physical analysis models. The application of deep learning models can capture subtle and complex nonlinear characteristics in vibration signals, while the integration of physical models provides a deep understanding of bearing fatigue behavior. In addition, one-dimensional deep feature extraction optimizes computational efficiency and reduces reliance on computing resources while maintaining sensitivity to subtle signal changes. This combined application ensures high prediction accuracy and rapid response capabilities under diverse industrial conditions.

[0124] This embodiment provides a life prediction method for an engine connecting rod big end bearing. Based on a fatigue analysis physical model of the engine connecting rod big end bearing, the physical model fatigue factor of the connecting rod big end bearing is determined, and the physical model fatigue factor is mapped to an initial category label; vibration signals at different times during the entire life cycle of the connecting rod big end bearing are obtained, and one-dimensional deep feature extraction is performed on the segmented vibration signal to obtain a feature vector corresponding to the vibration signal; a classification unit and a clustering unit are initialized based on the initial category label, the feature vector is input into the classification unit and the clustering unit, a loss function between the initial category label and the output category label is determined, and network parameters of the classification unit and the clustering unit are updated; dynamic weights are assigned based on the divergence difference between the initial category label and the category label; based on the dynamic weights, the initial category label is corrected, the loss function and the dynamic weights are re-determined, and the iteration is continued until the termination condition is met, thereby obtaining a target classification unit and a target clustering unit with the target initial category label; based on the output results of the target classification unit and the target clustering unit, a life prediction result of the connecting rod big end bearing is determined. Based on the physical analysis model, an initial life label is assigned to the collected vibration signal of the connecting rod big end bearing, and then the deep features of the vibration signal are extracted. Finally, based on the KL divergence, the life label is continuously and adaptively corrected during the model training process, ultimately obtaining a more stable and accurate life prediction result. Combining the vibration signal analysis of the deep learning model with the safety factor calculation of the physical model, the advantages of both are utilized to improve the accuracy and reliability of the prediction. Moreover, through multi-dimensional analysis, considering the actual operating conditions and material properties of the equipment, it can adapt to the life prediction needs under different working conditions and usage conditions, reducing the dependence on manual experience, and has good versatility and adaptability.

[0125] For example, in order to help understand the implementation process of the engine connecting rod big end bearing life prediction method obtained by combining this embodiment with the above embodiment 1, please refer to Figure 5 , Figure 5 A simplified flow chart of a method for predicting the life of an engine connecting rod big end bearing is provided. Specifically:

[0126] Use the physical model life analysis unit to obtain the physical model fatigue factor of the connecting rod big end bearing; collect vibration signals at different times during the entire life cycle of the connecting rod big end bearing, and extract the deep features of all vibration signals through the one-dimensional deep autoencoding feature unit; convert the obtained physical model fatigue factor into the corresponding bearing life label as the initial category label; initialize the classification unit and clustering unit according to the initial category label; calculate the loss function based on the initial category label and the category label output by the classification unit and clustering unit; use the KL divergence to calculate the difference between the category label and the initial category label, and assign dynamic weights accordingly to correct the initial category label; in the final stage of training, the network parameters can be fine-tuned to further optimize the classification performance; combine the dynamic weights and classifier output to obtain the final life prediction result.

[0127] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the engine connecting rod big end bearing life prediction method of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0128] This application also provides a device for predicting the life of the engine connecting rod big end bearing. Figure 6 , the engine connecting rod big end bearing life prediction device includes:

[0129] An analysis module 10 is configured to determine a physical model fatigue factor of the connecting rod big end bearing based on a fatigue analysis physical model of the engine connecting rod big end bearing, and map the physical model fatigue factor to an initial category label;

[0130] Correction module 20, for obtaining vibration signals of the connecting rod big end bearing at different times during its entire life cycle, performing one-dimensional deep feature extraction on the segmented vibration signals, and obtaining feature vectors corresponding to the vibration signals;

[0131] The correction module 20 is further used to initialize the classification unit and the clustering unit based on the initial category label, input the feature vector into the classification unit and the clustering unit, determine the loss function between the initial category label and the output category label, and update the network parameters of the classification unit and the clustering unit;

[0132] The correction module 20 is further configured to assign dynamic weights based on the divergence difference between the initial category labels and the category labels;

[0133] The correction module 20 is further used to correct the initial category label based on the dynamic weight, re-determine the loss function and the dynamic weight, and iterate continuously until the termination condition is met to obtain the target classification unit and the target clustering unit with the target initial category label;

[0134] The prediction module 30 is used to determine the life prediction result of the connecting rod big end bearing based on the output results of the target classification unit and the target clustering unit.

[0135] In a feasible embodiment, the analysis module 10 is further used to obtain fatigue influencing parameters of the connecting rod big end bearing, where the fatigue influencing parameters include at least local maximum stress, nominal stress, average stress, ultimate tensile strength of the bearing material, fatigue strength of a standard test piece, fatigue strength of an actual sample, surface fatigue strength of a smooth standard test piece, surface fatigue strength of an actual sample, symmetric fatigue limit of the bearing material, and sensitivity coefficient of the bearing material to stress cycle asymmetry;

[0136] Based on the first correspondence between fatigue influencing parameters and fatigue factors of the physical model, a fatigue analysis physical model is constructed.

[0137] In a feasible embodiment, the correction module 20 is also used to perform one-dimensional deep feature extraction on the segmented vibration signal through a one-dimensional deep self-encoding feature unit to obtain a feature vector corresponding to the vibration signal. The one-dimensional deep self-encoding feature unit includes an encoding network and a decoding network. The encoding network includes a convolution layer and a pooling layer, and the decoding network includes an upsampling layer and a convolution layer.

[0138] In a feasible embodiment, the category label includes a first category label and a second category label, and the correction module 20 is further used to input the feature vector into the classification unit and the clustering unit to obtain the first category label output by the classification unit and the second category label output by the clustering unit respectively;

[0139] Calculate the classification loss based on the initial category label and the first category label;

[0140] Calculate the clustering loss based on the initial category label and the second category label;

[0141] Determine the loss function based on classification loss and clustering loss;

[0142] Based on the gradient information of the loss function, the network parameters of the classification unit and clustering unit are updated.

[0143] In a feasible implementation, the correction module 20 is further configured to obtain the reconstruction loss of the vibration signal during the one-dimensional depth feature extraction process;

[0144] Obtain the second correspondence between classification loss, clustering loss, reconstruction loss, balance coefficient and loss function;

[0145] Based on the classification loss, clustering loss, reconstruction loss, balance coefficient and the second correspondence, a loss function between the initial category label and the category label is determined.

[0146] In a feasible implementation, the correction module 20 is further configured to determine a label mean based on the category label and the initial category label;

[0147] Obtain the third correspondence between the initial category label, label mean and divergence difference;

[0148] Based on the initial category labels, the label means, and the third correspondence, a divergence difference between the initial category labels and the category labels is determined.

[0149] In a feasible implementation manner, the correction module 20 is further configured to obtain a fourth corresponding relationship between the divergence difference, the adjustment parameter, and the dynamic weight;

[0150] A dynamic weight is determined based on a divergence difference between the initial category label and the category label, an adjustment parameter, and a fourth corresponding relationship.

[0151] In a feasible implementation, the correction module 20 is further configured to obtain a fifth correspondence between the initial category label, the category label, the category label weight, the dynamic weight, the transformation coefficient, and the corrected initial category label;

[0152] A revised initial category label is determined based on the initial category label, the category label, the weight of the category label, the dynamic weight, the transformation coefficient, and the fifth corresponding relationship.

[0153] The engine connecting rod big end bearing life prediction device provided in this application utilizes the engine connecting rod big end bearing life prediction method described in the aforementioned embodiment, thereby resolving the technical issue of low accuracy in engine connecting rod big end bearing life prediction. Compared to the prior art, the engine connecting rod big end bearing life prediction device provided in this application achieves the same beneficial effects as the engine connecting rod big end bearing life prediction method described in the aforementioned embodiment. Other technical features of the engine connecting rod big end bearing life prediction device are the same as those disclosed in the aforementioned embodiment and are not further elaborated upon here.

[0154] The present application provides an engine connecting rod big end bearing life prediction device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the engine connecting rod big end bearing life prediction method in the above-mentioned embodiment one.

[0155] Reference below Figure 5, which shows a schematic structural diagram of an engine connecting rod big end bearing life prediction device suitable for implementing an embodiment of the present application. The engine connecting rod big end bearing life prediction device in the embodiment of the present application can include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 7 The engine connecting rod big end bearing life prediction device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0156] like Figure 7 As shown, the engine connecting rod big end bearing life prediction device may include a processing device 1001 (e.g., a central processing unit, graphics processing unit, etc.), which can perform various appropriate actions and processes based on programs stored in a read-only memory (ROM) 1002 or programs loaded from a storage device 1003 into a random access memory (RAM) 1004. RAM 1004 also stores various programs and data required for the operation of the engine connecting rod big end bearing life prediction device. Processing device 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input device 1007 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output device 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage device 1003 including, for example, a magnetic tape, hard disk, etc.; and communication device 1009. Communication device 1009 can allow the engine connecting rod big end bearing life prediction device to communicate wirelessly or wired with other devices to exchange data. Although the figure shows an engine connecting rod big end bearing life prediction device with various systems, it should be understood that implementation or presence of all the illustrated systems is not required. More or fewer systems may alternatively be implemented or present.

[0157] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0158] The engine connecting rod big end bearing life prediction device provided in this application utilizes the engine connecting rod big end bearing life prediction method described in the aforementioned embodiment, thereby resolving the technical issue of low engine connecting rod big end bearing life prediction accuracy. Compared to the prior art, the engine connecting rod big end bearing life prediction device provided in this application achieves the same beneficial effects as the engine connecting rod big end bearing life prediction method described in the aforementioned embodiment. Other technical features of this engine connecting rod big end bearing life prediction device are the same as those disclosed in the aforementioned embodiment and are not further elaborated upon here.

[0159] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0160] The above are only specific embodiments of the present application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0161] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, a computer program) stored thereon, the computer-readable program instructions being used to execute the engine connecting rod big end bearing life prediction method in the above-mentioned embodiment.

[0162] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0163] The computer-readable storage medium may be included in the engine connecting rod big end bearing life prediction device; or may exist independently without being assembled into the engine connecting rod big end bearing life prediction device.

[0164] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the engine connecting rod big end bearing life prediction device, the engine connecting rod big end bearing life prediction device: determines the physical model fatigue factor of the connecting rod big end bearing based on the fatigue analysis physical model of the engine connecting rod big end bearing, and maps the physical model fatigue factor to an initial category label; obtains vibration signals at different times during the entire life cycle of the connecting rod big end bearing, performs one-dimensional deep feature extraction on the segmented vibration signals, and obtains a feature vector corresponding to the vibration signal; initializes the classification unit and the clustering unit based on the initial category label, inputs the feature vector into the classification unit and the clustering unit, determines the loss function between the initial category label and the output category label, and updates the network parameters of the classification unit and the clustering unit; assigns a dynamic weight based on the divergence difference between the initial category label and the category label; corrects the initial category label based on the dynamic weight, re-determines the loss function and the dynamic weight, and iterates continuously until a termination condition is met, thereby obtaining a target classification unit and a target clustering unit with a target initial category label; and determines the life prediction result of the connecting rod big end bearing based on the output results of the target classification unit and the target clustering unit.

[0165] Computer program code for performing the operations of the present application may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0166] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0167] The modules described in the embodiments of the present application may be implemented in software or hardware, wherein the name of a module does not necessarily limit the unit itself.

[0168] The computer-readable storage medium provided in this application stores computer-readable program instructions (i.e., a computer program) for executing the aforementioned engine connecting rod big end bearing life prediction method. This computer-readable storage medium can address the technical issue of low engine connecting rod big end bearing life prediction accuracy. Compared to the prior art, the computer-readable storage medium provided in this application offers the same beneficial effects as the engine connecting rod big end bearing life prediction method provided in the aforementioned embodiment, and will not be further elaborated here.

[0169] The present application also provides a computer program product, comprising a computer program, which implements the steps of the above-mentioned engine connecting rod big end bearing life prediction method when executed by a processor.

[0170] The computer program product provided in this application can address the technical issue of low accuracy in predicting the life of an engine connecting rod big end bearing. Compared to the prior art, the computer program product provided in this application offers the same beneficial effects as the engine connecting rod big end bearing life prediction method provided in the aforementioned embodiment, and will not be further elaborated here.

[0171] The above are only some embodiments of the present application and are not intended to limit the patent scope of the present application. All equivalent structural transformations made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A method for predicting the life of an engine connecting rod big end bearing, characterized in that: The method comprises: Determining a physical model fatigue factor of the connecting rod big end bearing based on a fatigue analysis physical model of the engine connecting rod big end bearing, and mapping the physical model fatigue factor to an initial category label; Acquire vibration signals of the connecting rod big end bearing at different times throughout its life cycle, perform one-dimensional deep feature extraction on the segmented vibration signals, and obtain feature vectors corresponding to the vibration signals; Initializing a classification unit and a clustering unit based on the initial category label, inputting the feature vector into the classification unit and the clustering unit, determining a loss function between the initial category label and the output category label, and updating network parameters of the classification unit and the clustering unit; assigning a dynamic weight based on a divergence difference between the initial category label and the category label; Based on the dynamic weight, the initial category label is modified, the loss function and the dynamic weight are re-determined, and the iteration is continued until the termination condition is met, thereby obtaining a target classification unit and a target clustering unit with the target initial category label; Determining a life prediction result of the connecting rod big end bearing based on output results of the target classification unit and the target clustering unit; Before the step of determining the fatigue factor of the physical model of the connecting rod big end bearing based on the fatigue analysis physical model of the engine connecting rod big end bearing, the method further includes: Obtaining fatigue influencing parameters of the connecting rod big end bearing, the fatigue influencing parameters including at least local maximum stress, nominal stress, average stress, ultimate tensile strength of the bearing material, fatigue strength of a standard test piece, fatigue strength of an actual sample, surface fatigue strength of a smooth standard test piece, surface fatigue strength of an actual sample, symmetric fatigue limit of the bearing material, and sensitivity coefficient of the bearing material to stress cycle asymmetry; The fatigue analysis physical model is constructed based on a first corresponding relationship between fatigue influencing parameters and physical model fatigue factors.

2. The method according to claim 1, wherein The step of performing one-dimensional deep feature extraction on the segmented vibration signal to obtain a feature vector corresponding to the vibration signal includes: The segmented vibration signal is subjected to one-dimensional deep feature extraction through a one-dimensional deep self-encoding feature unit to obtain a feature vector corresponding to the vibration signal. The one-dimensional deep self-encoding feature unit includes an encoding network and a decoding network. The encoding network includes a convolution layer and a pooling layer, and the decoding network includes an upsampling layer and a convolution layer.

3. The method according to claim 1, wherein The category label includes a first category label and a second category label, and the steps of inputting the feature vector into the classification unit and the clustering unit, determining a loss function between the initial category label and the output category label, and updating network parameters of the classification unit and the clustering unit include: Inputting the feature vector into the classification unit and the clustering unit to obtain a first category label output by the classification unit and a second category label output by the clustering unit respectively; Calculating a classification loss based on the initial category label and the first category label; Calculating a clustering loss based on the initial category label and the second category label; Determining the loss function based on the classification loss and the clustering loss; Based on the gradient information of the loss function, the network parameters of the classification unit and the clustering unit are updated.

4. The method according to claim 3, wherein The step of determining the loss function based on the classification loss and the clustering loss includes: Obtaining a reconstruction loss of the vibration signal during a one-dimensional deep feature extraction process; Obtain the second correspondence between classification loss, clustering loss, reconstruction loss, balance coefficient and loss function; A loss function between the initial category label and the category label is determined based on the classification loss, the clustering loss, the reconstruction loss, a balance coefficient, and the second corresponding relationship.

5. The method according to claim 1, wherein Before the step of allocating a dynamic weight based on the divergence difference between the initial category label and the category label, the step further includes: Determining a label mean based on the category label and the initial category label; Obtain the third correspondence between the initial category label, label mean and divergence difference; Based on the initial category label, the label mean, and the third corresponding relationship, a divergence difference between the initial category label and the category label is determined.

6. The method according to claim 1, wherein The step of allocating a dynamic weight based on the divergence difference between the initial category label and the category label comprises: Obtaining a fourth correspondence between the divergence difference, the adjustment parameter, and the dynamic weight; The dynamic weight is determined based on a divergence difference between the initial category label and the category label, an adjustment parameter, and the fourth corresponding relationship.

7. The method according to claim 1, wherein The step of correcting the initial category label based on the dynamic weight includes: Obtaining a fifth correspondence between the initial category label, the category label, the category label weight, the dynamic weight, the transformation coefficient, and the revised initial category label; A revised initial category label is determined based on the initial category label, the category label, the weight of the category label, the dynamic weight, the transformation coefficient, and the fifth corresponding relationship.

8. A device for predicting the life of an engine connecting rod big end bearing, characterized in that: The device comprises: an analysis module, configured to determine a physical model fatigue factor of the connecting rod big end bearing based on a fatigue analysis physical model of the connecting rod big end bearing, and map the physical model fatigue factor to an initial category label; a correction module, configured to obtain vibration signals of the connecting rod big end bearing at different times during the entire life cycle, perform one-dimensional deep feature extraction on the segmented vibration signals, and obtain a feature vector corresponding to the vibration signal; The correction module is further configured to initialize a classification unit and a clustering unit based on the initial category label, input the feature vector into the classification unit and the clustering unit, determine a loss function between the initial category label and the output category label, and update network parameters of the classification unit and the clustering unit; The correction module is further configured to assign a dynamic weight based on a divergence difference between the initial category label and the category label; The correction module is further configured to correct the initial category label based on the dynamic weight, redetermine the loss function and the dynamic weight, and iterate continuously until a termination condition is met, thereby obtaining a target classification unit and a target clustering unit with a target initial category label; a prediction module, configured to determine a life prediction result of the connecting rod big end bearing based on output results of the target classification unit and the target clustering unit; The analysis module is further configured to obtain fatigue influencing parameters of the connecting rod big end bearing, wherein the fatigue influencing parameters include at least local maximum stress, nominal stress, average stress, ultimate tensile strength of the bearing material, fatigue strength of a standard test piece, fatigue strength of an actual sample, surface fatigue strength of a smooth standard test piece, surface fatigue strength of an actual sample, symmetric fatigue limit of the bearing material, and sensitivity coefficient of the bearing material to stress cycle asymmetry; The fatigue analysis physical model is constructed based on a first corresponding relationship between fatigue influencing parameters and physical model fatigue factors.

9. An engine connecting rod big end bearing life prediction device, characterized in that: The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the engine connecting rod big end bearing life prediction method according to any one of claims 1 to 7.

Images