Bearing residual life interval prediction method and system

Abstract

The application discloses a bearing residual life interval prediction method and system, and the method comprises the following steps: collecting signal data of a bearing at time t, and predicting a life prediction value of the bearing at time t based on the signal data; predicting an upper limit of a residual life interval of the bearing at time t based on bearing life cycle data stored in a historical knowledge base and the life prediction value of the bearing at time t; predicting a lower limit of the residual life interval of the bearing at time t based on the life prediction value of the bearing at time t and complete life cycle data or error data based on signal data at time t and signal data at historical time; and determining the residual life interval of the bearing at time t based on the upper limit of the residual life interval of the bearing at time t and the lower limit of the residual life interval of the bearing at time t.

Description

Technical Field

[0001] This invention relates to the field of equipment health management technology, specifically to a method and system for predicting the remaining life range of bearings. Background Technology

[0002] In the battery manufacturing process, slurry mixing is the most upstream and one of the most important processes. The core of the slurry mixing motor is the bearing. Accurate prediction of the bearing life of the slurry mixing motor can enable timely maintenance of the equipment, avoid production stoppages caused by mechanical failures and wear, and guide manual maintenance and spare parts in a timely manner, transforming abnormal maintenance into preventive maintenance.

[0003] Remaining useful life (RUL) prediction research includes methods based on equipment degradation mechanisms, data-driven statistical and machine learning methods, and methods that combine mechanisms and statistics. However, current RUL prediction research basically directly applies various deep networks proposed and developed under the application needs of other fields. Their network structures and parameters are deterministic, and they can generally only obtain deterministic RUL prediction values. However, in actual use, predicting only fixed values ​​is quite limiting, and the ability to quantify the uncertainty in prediction is obviously insufficient.

[0004] Existing interval prediction methods can be broadly categorized into two types: statistical methods, which use statistical techniques to estimate the interval; and loss function methods, which output the interval by defining a specific loss function. In recent years, depth-based interval prediction methods have primarily focused on loss functions, such as quantile loss. These methods apply different penalty weights to overfitting and underfitting predictions to obtain upper and lower bounds for the predicted values. However, this approach only provides a seemingly reasonable upper and lower bound for the prediction; in reality, the model does not learn about unknown factors at the current moment, resulting in inaccurate predictions of the remaining lifespan interval.

[0005] In related technologies, Chinese invention patent application CN112818870A discloses a method for predicting the remaining life of bearings based on a gated neural network framework. This method uses the vibration signal of the tested bearing as input, and the output of the gated neural network model is the remaining life of the tested bearing at the current moment. By adding a Bayesian layer, the traditional point prediction result is transformed into an interval prediction to account for the uncertainty in bearing remaining life prediction. However, this scheme uses a GRU + attention mechanism for point prediction and a Bayesian algorithm for probability distribution fitting and prediction to achieve interval prediction. The interval prediction requires model training, making the data processing complex. Furthermore, although the Bayesian layer learns the probability distribution of the interval, it still does not explicitly express the fluctuation of life under different sensor data fluctuations.

[0006] Chinese invention patent application CN115098960A discloses a method and apparatus for predicting the remaining useful life of equipment. The method involves calculating the similarity between currently obtained equipment data and various full-lifetime data in a historical database, selecting multiple sets of historical data whose similarity meets preset conditions; calculating the similarity between data under different operating conditions in the currently obtained equipment data and data under different operating conditions in the multiple sets of historical data meeting the preset conditions, obtaining the position and similarity score with the highest similarity; calculating the remaining useful life under different operating conditions based on the obtained position and similarity score, and determining the equipment's remaining useful life point prediction result based on the subsequent task profile of the equipment; and determining the interval prediction result of the equipment's remaining useful life based on the equipment's remaining useful life point prediction result.

[0007] However, this scheme uses similarity for point estimation; in the process of predicting the remaining life interval, the upper limit data and lower limit data training set of the interval are constructed by the residual between the model output prediction result of the training sample and the true value of the standard output to train the model and make predictions, thereby obtaining the upper limit and lower limit estimates of the interval; it can be seen that the calculation of the upper and lower limits in this scheme uses the predicted value to make interval predictions, which itself has certain problems and errors. Summary of the Invention

[0008] The technical problem to be solved by this invention is how to improve the accuracy of bearing remaining life range prediction.

[0009] The present invention solves the above-mentioned technical problems through the following technical means:

[0010] On the one hand, a method for predicting the remaining life range of bearings is proposed, the method comprising:

[0011] Collect bearings t The signal data at a given time, and the bearing prediction based on the signal data. t Lifespan prediction at any given time;

[0012] Based on the bearing lifecycle data stored in the historical knowledge base and the bearing t Predicted lifespan value at any time, predicting bearing lifespan. t The upper limit of the remaining lifetime range at any given moment;

[0013] Based on bearings t The predicted lifetime value at a given moment and the complete lifecycle data or based on t Error data between the current time signal data and its historical time signal data are used to predict bearings. t Lower limit of the remaining lifetime interval at any given time

[0014] Based on the bearing t The upper limit of the remaining lifespan range at any given time and the bearing t The lower limit of the remaining lifespan range at any given time determines the bearing's position. t The remaining lifespan at any given moment.

[0015] Furthermore, the acquisition bearing t The signal data at a given time, and the bearing prediction based on the signal data. t The predicted lifetime values ​​at each moment include:

[0016] Collect the bearing t The signal data at any given time, including multi-source signal data and lifetime;

[0017] The bearing t The signal data at each moment is used as input to the life prediction model to predict the bearing's lifespan. t The predicted lifetime value at time t, wherein the lifetime prediction model adopts a structure combining a convolutional neural network and a long short-term memory network.

[0018] Furthermore, in the process of placing the bearing t Before using the signal data at a given time as input to the lifetime prediction model, the method further includes:

[0019] For the bearing t The signal data at each time point is preprocessed to obtain the data format that serves as the input to the lifetime prediction model. The preprocessing methods include smoothing missing data, segmenting the data, normalizing the data, and combining features.

[0020] Furthermore, the complete lifecycle data includes historical multi-source signal data of the bearing at each moment and historical remaining life at each moment, based on the bearing lifecycle data stored in the historical knowledge base and the bearing... t Predicted lifespan value at any time, predicting bearing lifespan. t The upper limit of the remaining lifetime interval at any given time includes:

[0021] Obtain historical data from the historical knowledge base that indicates the bearing has a remaining lifespan greater than that of the bearing. t The historical multi-source signal data corresponding to the lifetime prediction value at each moment;

[0022] Calculate each historical multi-source signal data and the bearing separately. t Similarity of multi-source signal data at any given time;

[0023] Calculate the similarity of high n The first average of the remaining lifetime corresponding to each historical multi-source signal data point;

[0024] Based on the first average of the historical remaining life and the bearing t Predicted lifespan value at any time, predicting bearing lifespan. t The upper limit of the remaining lifespan range at any given time.

[0025] Furthermore, the bearing t The formula for predicting the upper limit of the remaining lifetime interval at a given time is:

[0026]

[0027] In the formula: For bearings t The upper limit of the remaining lifetime range at any given moment. For bearings t Life expectancy prediction at any given time. This represents the first average of historical remaining lifespan.

[0028] Furthermore, the complete lifecycle data includes historical multi-source signal data of the bearing at each moment and historical remaining life at each moment, based on the bearing. t Based on the predicted lifespan value at a given moment and the complete lifespan data, the bearing is predicted. t When the lower limit of the remaining lifetime interval is reached, it includes:

[0029] Obtain from the historical knowledge base that the bearing's remaining lifespan is less than that of the bearing. t The historical multi-source signal data corresponding to the lifetime prediction value at each moment;

[0030] Calculate each historical multi-source signal data and the bearing separately. t Similarity of multi-source signal data at any given time;

[0031] Calculate the similarity of high n The second average of the remaining lifetime corresponding to each historical multi-source signal data point;

[0032] Based on the second average of the historical remaining life and the bearing t Predicted lifespan value at any time, predicting bearing lifespan.t The lower limit of the remaining lifetime range at any given time.

[0033] Furthermore, the bearing t The formula for predicting the lower limit of the remaining lifetime interval at a given time is:

[0034]

[0035] In the formula: For bearings t The lower limit of the remaining lifetime range at any given time. For bearings t Life expectancy prediction at any given time. This is the second average of the historical remaining lifespan.

[0036] Furthermore, based on t Error data between the current time signal data and its historical time signal data are used to predict bearings. t When the lower limit of the remaining lifetime interval is reached, it includes:

[0037] Calculate the bearing t Time signal data and t The difference between the signal data at time -1 is used as the error data;

[0038] The error data is used as input to the error prediction model to predict bearing error. t The lifetime error at time point, wherein the error prediction model employs a long short-term memory network;

[0039] Based on the bearing t The predicted lifespan value at a given time and the lifespan error are used to predict the bearing's lifespan. t The lower limit of the remaining lifetime range at any given time.

[0040] Furthermore, the bearing-based t The predicted lifespan value at a given time and the lifespan error are used to predict the bearing's lifespan. t The lower limit of the remaining lifetime interval at any given time is expressed by the formula:

[0041]

[0042] In the formula: For bearings t The lower limit of the remaining lifetime range at any given time. For bearings t Life expectancy prediction at any given time. This refers to the lifetime error.

[0043] On the other hand, a bearing remaining life range prediction system is also proposed, the system comprising:

[0044] Data acquisition module, used to collect bearing data.t The signal data at a given time, and the bearing prediction based on the signal data. t Lifespan prediction at any given time;

[0045] The interval upper limit prediction module is used to predict bearing lifecycle data stored in the historical knowledge base and the bearing. t Predicted lifespan value at any time, predicting bearing lifespan. t The upper limit of the remaining lifetime range at any given moment;

[0046] The lower limit prediction module for the interval is used for bearing-based prediction. t The predicted lifetime value at a given moment and the complete lifecycle data or based on t Error data between the current time signal data and its historical time signal data are used to predict bearings. t The lower limit of the remaining lifetime range at any given time;

[0047] The remaining lifespan prediction module is used to predict the remaining lifespan based on the bearing. t The upper limit of the remaining lifespan range at any given time and the bearing t The lower limit of the remaining lifespan range at any given time determines the bearing's position. t The remaining lifespan at any given moment.

[0048] The advantages of this invention are:

[0049] (1) This invention is based on bearings t The bearing's lifespan prediction value at a given moment is compared with the bearing lifespan data in the historical knowledge base. t Predicting the upper limit of the remaining lifespan range at any given moment, based on bearings, when the knowledge base contains complete lifespan data for the device. t The bearing's lifespan prediction at any given moment is compared with the actual bearing lifespan data in the historical knowledge base. t When the knowledge base lacks complete life cycle data for the device, the lower limit of the remaining life cycle range is predicted. However, a neural network is used to predict the lower limit error range. The range is estimated by using the bearing's historical life cycle data, incorporating prior knowledge. At the same time, a neural network is introduced to predict and fit the range error based on historical values. This not only amplifies or reduces the predicted value, but also obtains an accurate bearing remaining life cycle range.

[0050] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0051] Figure 1 This is a schematic flowchart of a bearing remaining life range prediction method proposed in an embodiment of the present invention;

[0052] Figure 2This is a schematic diagram of the overall process of the bearing remaining life range prediction method proposed in an embodiment of the present invention;

[0053] Figure 3 This is an example diagram of data collection in one embodiment of the present invention;

[0054] Figure 4 This is a schematic diagram of the life prediction model in one embodiment of the present invention;

[0055] Figure 5 This is a principle block diagram of a bearing remaining life range prediction method proposed in an embodiment of the present invention;

[0056] Figure 6 This is a block diagram illustrating the principle of using a neural network to predict the lower limit error range in one embodiment of the present invention.

[0057] Figure 7 This is a schematic diagram of the bearing remaining life range prediction system proposed in an embodiment of the present invention. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] like Figures 1 to 2 As shown, the first embodiment of the present invention proposes a method for predicting the remaining life range of a bearing, the method comprising the following steps:

[0060] S10, Acquisition Bearing t The signal data at a given time, and the bearing prediction based on the signal data. t Lifespan prediction at any given time;

[0061] S20. Based on the bearing lifecycle data stored in the historical knowledge base and the bearing... t Predicted lifespan value at any time, predicting bearing lifespan. t The upper limit of the remaining lifetime range at any given moment;

[0062] S30, based on bearing t The predicted lifetime value at a given moment and the complete lifecycle data or based on t Error data between the current time signal data and its historical time signal data are used to predict bearings. t The lower limit of the remaining lifetime range at any given time;

[0063] S40, Based on the bearingt The upper limit of the remaining lifespan range at any given time and the bearing t The lower limit of the remaining lifespan range at any given time determines the bearing's position. t The remaining lifespan at any given moment.

[0064] This embodiment is based on bearings. t The bearing's lifespan prediction value at a given moment is compared with the bearing lifespan data in the historical knowledge base. t Predicting the upper limit of the remaining lifespan range at any given moment, based on bearings, when the knowledge base contains complete lifespan data for the device. t The bearing's lifespan prediction at any given moment is compared with the actual bearing lifespan data in the historical knowledge base. t When the knowledge base lacks complete life cycle data for the device, the lower limit of the remaining life cycle range is predicted. However, a neural network is used to predict the lower limit error range. The range is estimated by using the bearing's historical life cycle data, incorporating prior knowledge. At the same time, a neural network is introduced to predict and fit the range error based on historical values. This not only amplifies or reduces the predicted value, but also obtains an accurate bearing remaining life cycle range.

[0065] In one embodiment, step S10: collecting bearings t The signal data at a given time, and the bearing prediction based on the signal data. t The predicted lifetime value at a given time includes the following steps:

[0066] S11, Collect the bearing t The signal data at any given time, including multi-source signal data;

[0067] Specifically, multi-source signal data includes voltage and current signals, sound signals, force signals, temperature signals, and vibration signals, etc. An example of data acquisition is shown below. Figure 3 As shown, then The input at any given time consists of data collected by S sensors:

[0068]

[0069] In the formula: represent At this moment The data monitored by each sensor, S represents the total number of sensors, and T represents the total time.

[0070] S12, the bearing t The signal data at each moment is used as input to the life prediction model to predict the bearing's lifespan. t The predicted lifetime value at time t, wherein the lifetime prediction model adopts a structure combining a convolutional neural network and a long short-term memory network.

[0071] Specifically, the lifetime prediction model used in this embodiment is a pre-trained neural network model for predicting remaining lifetime based on multi-source signal data. During model training, the input is specified. The dimension is (batch size, time steps, feature numbers), input It can be expressed as , Representing the The first sample At the [time]th moment The input has 1 feature dimension, where , , Output label values The dimension is , representing the remaining lifetime of multiple samples at each time point, output It can be expressed as ,in , ; Representing the The first sample The remaining lifespan is used as a label for model training to calculate the loss between the model's predictions and the actual results, thereby updating the model parameters.

[0072] Since CNNs are better at extracting signal data, while LSTMs are better at processing time-series data, this embodiment uses the following... Figure 4 The CNN+LSTM structure shown is used for lifetime prediction. The lifetime prediction at each time point depends on the sensor data of the current node and the state data left at the previous time point. Compared with the GRU+attention mechanism for point prediction, the CNN+LSTM model used in this embodiment can learn linear and nonlinear features and has a better effect on processing signal data.

[0073] Among them, the The first sample The lifetime value at each time point can be obtained from the prediction model.

[0074]

[0075] In the formula: yes A combination of features, i.e. .

[0076] Therefore, according to the first Sample Real-time collected signal data and historical data The lifespan prediction model can be used to obtain the first... Sample Life expectancy prediction at time .

[0077] In one embodiment, such as Figure 5 As shown, in step S21: the bearing t Before using the signal data at a given time as input to the lifetime prediction model, the method further includes the following steps:

[0078] For the bearing t The signal data at each time point is preprocessed to obtain the data format that serves as the input to the lifetime prediction model. The preprocessing methods include smoothing missing data, segmenting the data, normalizing the data, and combining features.

[0079] It should be noted that this embodiment smooths the missing data. If there are too many sample points, the data needs to be segmented, then normalized and combined with features, and finally converted into the form of model input and output.

[0080] In one embodiment, the complete lifecycle data includes historical multi-source signal data of the bearing at each moment and historical remaining life at each moment. The bearing lifecycle data stored in the historical knowledge base and the bearing in step S20 are... t Predicted lifespan value at any time, predicting bearing lifespan. t The upper limit of the remaining lifetime interval at any given time includes the following steps:

[0081] S21. Obtain from the historical knowledge base a bearing with a remaining lifespan greater than that of the bearing. t The historical multi-source signal data corresponding to the lifetime prediction value at each moment;

[0082] Specifically, the complete lifecycle data of the device is stored in a historical knowledge base, with the following storage format: Time signal data Lifetime at time t Obtain lifespan-related information from historical knowledge bases. > In the N signal data Time of the first Signal data of one sample Therefore, calculate with "Similar" n Equipment data.

[0083] S22. Calculate the relationship between each historical multi-source signal data and the bearing. tSimilarity of multi-source signal data at any given time;

[0084] Specifically, this embodiment uses a vector metric to calculate the similarity between two signal data points, by calculating the input... With knowledge base Distance between various signal data The upper limit of the interval is estimated by averaging the lifetime records of the n closest data points. The similarity calculation formula compares the input features of the i-th sample at time t with all data in the knowledge base to find N similar data points. The formula is as follows:

[0085]

[0086] In the formula: express t In the context of time, the historical knowledge base is related to the first i N similar data points for each sample; express t Under time, the first i Input features of each sample; This represents all features of the knowledge base at all times from 0 to T; Representing the The first sample Multiple feature dimensions at a given time; The first in the knowledge base At the [time]th moment Features collected by each sensor.

[0087] It should be noted that this embodiment combines Euclidean distance and Manhattan distance to select Topn, with n defaulting to 10.

[0088] S23, Calculate the similarity of high-value items n The first average of the remaining lifetime corresponding to each historical multi-source signal data point;

[0089] S24, Based on the first average value of the historical remaining life and the bearing t Predicted lifespan value at any time, predicting bearing lifespan. t The upper limit of the remaining lifespan range at any given time.

[0090] In one embodiment, the bearing t The formula for predicting the upper limit of the remaining lifetime interval at a given time is:

[0091]

[0092] In the formula: For bearings t The upper limit of the remaining lifetime range at any given moment. For bearings t Life expectancy prediction at any given time. This represents the first average of historical remaining lifespan.

[0093] In one embodiment, in step S30, based on the bearing t Based on the predicted lifespan value at a given moment and the complete lifespan data, the bearing is predicted. t The lower limit of the remaining lifetime range at any given time includes the following steps:

[0094] S31. Obtain from the historical knowledge base the bearing with a remaining lifespan less than that of the bearing. t The historical multi-source signal data corresponding to the lifetime prediction value at each moment;

[0095] S32. Calculate the relationship between each historical multi-source signal data and the bearing. t Similarity of multi-source signal data at any given time;

[0096] S33, Calculate the similarity of high-value items n The second average of the remaining lifetime corresponding to each historical multi-source signal data point;

[0097] S34, Based on the second average value of the historical remaining life and the bearing t Predicted lifespan value at any time, predicting bearing lifespan. t The lower limit of the remaining lifetime range at any given time.

[0098] It should be noted that this embodiment obtains the lifespan from a historical knowledge base. And the signal data and The signal data at that moment were "similar". The equipment data is used to estimate the lower limit of the predicted value. Value ,middle: For bearings t The lower limit of the remaining lifetime range at any given time. For bearings t Life expectancy prediction at any given time. This is the second average of the historical remaining lifespan.

[0099] This embodiment estimates and predicts the range of predicted values ​​by borrowing a known historical knowledge base. It uses known information from the knowledge base to represent unknown range information, thereby improving the accuracy of bearing remaining life range prediction.

[0100] In one embodiment, in step S30, based on t Error data between the current time signal data and its historical time signal data are used to predict bearings. t The lower limit of the remaining lifetime interval at any given time includes the following steps:

[0101] S31', Calculate the bearing t Time signal data and t The difference between the signal data at time -1 is used as the error data;

[0102] S32'. Use the error data as input to the error prediction model to predict the bearing error. t The lifetime error at time point, wherein the error prediction model employs a long short-term memory network;

[0103] S33', based on the bearing t The predicted lifespan value at a given time and the lifespan error are used to predict the bearing's lifespan. t The lower limit of the remaining lifetime range at any given time.

[0104] It should be noted that this embodiment uses... Time data and Input error data is constructed by subtracting the time signal data. The predicted label is the next time. and The difference in lifetime at any given moment. The lifetime error at the next moment is predicted using error data from the current moment and its historical moments. Since the lifetime error is a function that decreases with machine runtime, the error is negatively evaluated, allowing the error prediction model to predict a value greater than or equal to 0.

[0105] It should be noted that, as Figure 6 As shown, the error prediction model used in this embodiment is a pre-trained model that can be used to predict lifetime errors. Specifically, LSTM is used to learn the error, ensuring that the error is non-negative in the future. Finally, the ReLU function is used to output the prediction error. ,according to The interval error at the next time step is inferred from the error value at time step one. .

[0106] In one embodiment, step S33': based on the bearing t The predicted lifespan value at a given time and the lifespan error are used to predict the bearing's lifespan. t The lower limit of the remaining lifetime interval at any given time is expressed by the formula:

[0107]

[0108] In the formula: For bearings t The lower limit of the remaining lifetime range at any given time. For bearings t Life expectancy prediction at any given time. This refers to the lifetime error.

[0109] This embodiment obtains the predicted value range through upper and lower bound estimation. When the knowledge base data is complete, it uses point prediction results and real values ​​similar to the input in the knowledge base for interval estimation. When the dataset is incomplete, it uses the lifetime value at a future time as the lower bound interval for the current time for estimation and learning. Specifically, it uses the error data between the current time data and its historical time data as input features, and uses the lifetime error between the current time and the next time as the training target to obtain the lower bound error estimate for the current time. It predicts the upper and lower bounds by using the real values. Remaining lifespan range. This embodiment uses real historical values ​​and real future errors to predict the range, resulting in more accurate data. By directly using historical data for estimation, it cleverly uses historical knowledge to express the fluctuation range of lifespan, thereby guiding the maintenance of industrial equipment.

[0110] In addition, such as Figure 7 As shown, the second embodiment of the present invention proposes a bearing remaining life range prediction system, the system comprising:

[0111] Data acquisition module 10 is used to collect bearing data. t The signal data at a given time, and the bearing prediction based on the signal data. t Lifespan prediction at any given time;

[0112] The interval upper limit prediction module 20 is used to predict the bearing life cycle data stored in the historical knowledge base and the bearing. t Predicted lifespan value at any time, predicting bearing lifespan. t The upper limit of the remaining lifetime range at any given moment;

[0113] The lower limit prediction module 30 is used for bearing-based prediction. t The predicted lifetime value at a given moment and the complete lifecycle data or based on t Error data between the current time signal data and its historical time signal data are used to predict bearings. t The lower limit of the remaining lifetime range at any given time;

[0114] The remaining lifespan prediction module 40 is used to predict the remaining lifespan based on the bearing. t The upper limit of the remaining lifespan range at any given time and the bearing t The lower limit of the remaining lifespan range at any given time determines the bearing's position. t The remaining lifespan at any given moment.

[0115] This embodiment is based on bearings. t The bearing's lifespan prediction value at a given moment is compared with the bearing lifespan data in the historical knowledge base. t Predicting the upper limit of the remaining lifespan range at any given moment, based on bearings, when the knowledge base contains complete lifespan data for the device. tThe bearing's lifespan prediction value at a given moment is compared with the bearing lifespan data in the historical knowledge base. t When the knowledge base lacks complete life cycle data for the device, the lower limit of the remaining life cycle range is predicted. However, a neural network is used to predict the lower limit error range. The range is estimated by using the bearing's historical life cycle data, incorporating prior knowledge. At the same time, a neural network is introduced to predict and fit the range error based on historical values. This not only amplifies or reduces the predicted value, but also obtains an accurate bearing remaining life cycle range.

[0116] In one embodiment, the data acquisition module 10 includes:

[0117] The acquisition unit is used to acquire data about the bearing. t The signal data at any given time, including multi-source signal data;

[0118] The life point prediction unit is used to predict the bearing's lifespan. t The signal data at each moment is used as input to the life prediction model to predict the bearing's lifespan. t The predicted lifetime value at time t, wherein the lifetime prediction model adopts a structure combining a convolutional neural network and a long short-term memory network.

[0119] In one embodiment, the system further includes a data preprocessing module, specifically used to perform the following steps:

[0120] For the bearing t The signal data at each time point is preprocessed to obtain the data format that serves as the input to the lifetime prediction model. The preprocessing methods include smoothing missing data, segmenting the data, normalizing the data, and combining features.

[0121] In one embodiment, the interval upper limit prediction module 20 is specifically used to perform the following steps:

[0122] Obtain historical data from the historical knowledge base that indicates the bearing has a remaining lifespan greater than that of the bearing. t The historical multi-source signal data corresponding to the lifetime prediction value at each moment;

[0123] Calculate each historical multi-source signal data and the bearing separately. t Similarity of multi-source signal data at any given time;

[0124] Calculate the similarity of high n The first average of the remaining lifetime corresponding to each historical multi-source signal data point;

[0125] Based on the first average of the historical remaining life and the bearing t Predicted lifespan value at any time, predicting bearing lifespan. tThe upper limit of the remaining lifespan range at any given time.

[0126] Furthermore, the bearing t The formula for predicting the upper limit of the remaining lifetime interval at a given time is:

[0127]

[0128] In the formula: For bearings t The upper limit of the remaining lifetime range at any given moment. For bearings t Life expectancy prediction at any given time. This represents the first average of historical remaining lifespan.

[0129] In one embodiment, the lower limit prediction module 30 is specifically used to perform the following steps:

[0130] Obtain from the historical knowledge base that the bearing's remaining lifespan is less than that of the bearing. t The historical multi-source signal data corresponding to the lifetime prediction value at each moment;

[0131] Calculate each historical multi-source signal data and the bearing separately. t Similarity of multi-source signal data at any given time;

[0132] Calculate the similarity of high n The second average of the remaining lifetime corresponding to each historical multi-source signal data point;

[0133] Based on the second average of the historical remaining life and the bearing t Predicted lifespan value at any time, predicting bearing lifespan. t The lower limit of the remaining lifetime range at any given time.

[0134] Furthermore, the bearing t The formula for predicting the lower limit of the remaining lifetime interval at a given time is:

[0135]

[0136] In the formula: For bearings t The lower limit of the remaining lifetime range at any given time. For bearings t Life expectancy prediction at any given time. This is the second average of the historical remaining lifespan.

[0137] In one embodiment, the lower limit prediction module 30 is further configured to perform the following steps:

[0138] Calculate the bearing t Time signal data and tThe difference between the signal data at time -1 is used as the error data;

[0139] The error data is used as input to the error prediction model to predict bearing error. t The lifetime error at time point, wherein the error prediction model employs a long short-term memory network;

[0140] Based on the bearing t The predicted lifespan value at a given time and the lifespan error are used to predict the bearing's lifespan. t The lower limit of the remaining lifetime range at any given time.

[0141] Furthermore, the bearing-based t The predicted lifespan value at a given time and the lifespan error are used to predict the bearing's lifespan. t The lower limit of the remaining lifetime interval at any given time is expressed by the formula:

[0142]

[0143] In the formula: For bearings t The lower limit of the remaining lifetime range at any given time. For bearings t Life expectancy prediction at any given time. This refers to the lifetime error.

[0144] It should be noted that other embodiments or implementation methods of the bearing remaining life range prediction system of the present invention can refer to the above-described method embodiments, and will not be repeated here.

[0145] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0146] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0147] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for predicting the remaining life range of a bearing, characterized in that, The method includes: Collect bearings t The signal data at a given time, and the bearing prediction based on the signal data. t Lifespan prediction at any given time; Based on the bearing lifecycle data stored in the historical knowledge base and the bearing t Predicted lifespan value at any time, predicting bearing lifespan. t The upper limit of the remaining lifespan range at any given time includes information obtained from the historical knowledge base regarding historical remaining lifespans greater than that of the bearing. t The predicted lifetime value at each moment corresponds to each historical multi-source signal data; the relationship between each historical multi-source signal data and the bearing is calculated. t Similarity of multi-source signal data at different times; calculation of data with high similarity n The first average of the historical remaining lifetime corresponding to the historical multi-source signal data; based on the first average of the historical remaining lifetime and the bearing t Predicted lifespan value at any time, predicting bearing lifespan. t The upper limit of the remaining lifespan range at any given time, wherein the bearing... t The formula for predicting the upper limit of the remaining lifetime interval at a given time is: In the formula: For bearings t The upper limit of the remaining lifetime range at any given moment. For bearings t Life expectancy prediction at any given time. This is the first average of the historical remaining lifespan; Based on bearings t Predicted lifespan values ​​at any given time and complete lifespan data or based on t Error data between the current time signal data and its historical time signal data are used to predict bearings. t The lower limit of the remaining lifespan interval at any given time, including the calculation of the bearing. t Time signal data and t The difference between the signal data at time -1 is used as error data; this error data is used as input to the error prediction model to predict the bearing. t The lifetime error at any given time, wherein the error prediction model employs a long short-term memory network; based on the bearing t The predicted lifespan value at a given time and the lifespan error are used to predict the bearing's lifespan. t The lower limit of the remaining lifetime interval at any given time is expressed by the formula: In the formula: For bearings t The lower limit of the remaining lifetime range at any given time. For bearings t Life expectancy prediction at any given time. This refers to the lifetime error; Based on the bearing t The upper limit of the remaining lifespan range at any given time and the bearing t The lower limit of the remaining lifespan range at any given time determines the bearing's position. t The remaining lifespan at any given moment.

2. The bearing remaining life range prediction method as described in claim 1, characterized in that, The acquisition bearing t The signal data at a given time, and the bearing prediction based on the signal data. t The predicted lifetime values ​​at each moment include: Collect the bearing t The signal data at any given time, including multi-source signal data; The bearing t The signal data at each moment is used as input to the life prediction model to predict the bearing's lifespan. t The predicted lifetime value at time t, wherein the lifetime prediction model adopts a structure combining a convolutional neural network and a long short-term memory network.

3. The bearing remaining life range prediction method as described in claim 2, characterized in that, The bearing t Before using the signal data at a given time as input to the lifetime prediction model, the method further includes: For the bearing t The signal data at each time point is preprocessed to obtain the data format that serves as the input to the lifetime prediction model. The preprocessing methods include smoothing missing data, segmenting the data, normalizing the data, and combining features.

4. The bearing remaining life range prediction method as described in claim 1, characterized in that, The complete lifecycle data includes historical multi-source signal data of the bearing at each moment and historical remaining life at each moment, based on the bearing. t Based on the predicted lifespan value at a given moment and the complete lifespan data, the bearing is predicted. t When the lower limit of the remaining lifetime interval is reached, it includes: Obtain from the historical knowledge base that the bearing's remaining lifespan is less than that of the bearing. t The historical multi-source signal data corresponding to the lifetime prediction value at each moment; Calculate each historical multi-source signal data and the bearing separately. t Similarity of multi-source signal data at any given time; Calculate the similarity of high n The second average of the remaining lifetime corresponding to each historical multi-source signal data point; Based on the second average of the historical remaining life and the bearing t Predicted lifespan value at any time, predicting bearing lifespan. t The lower limit of the remaining lifetime range at any given time.

5. The bearing remaining life range prediction method as described in claim 4, characterized in that, The bearing t The formula for predicting the lower limit of the remaining lifetime interval at a given time is: In the formula: For bearings t The lower limit of the remaining lifetime range at any given time. For bearings t Life expectancy prediction at any given time. This is the second average of the historical remaining lifespan.

6. A bearing remaining life range prediction system, characterized in that, For implementing the bearing remaining life range prediction method as described in any one of claims 1-5, the system comprises: Data acquisition module, used to collect bearing data. t The signal data at a given time, and the bearing prediction based on the signal data. t Lifespan prediction at any given time; The interval upper limit prediction module is used to predict bearing lifecycle data stored in the historical knowledge base and the bearing. t Predicted lifespan value at any time, predicting bearing lifespan. t The upper limit of the remaining lifetime range at any given moment; The lower limit prediction module for the interval is used for bearing-based prediction. t Predicted lifespan values ​​at any given time and complete lifespan data or based on t Error data between the current time signal data and its historical time signal data are used to predict bearings. t The lower limit of the remaining lifetime range at any given time; The remaining lifespan prediction module is used to predict the remaining lifespan based on the bearing. t The upper limit of the remaining lifespan range at any given time and the bearing t The lower limit of the remaining lifespan range at any given time determines the bearing's position. t The remaining lifespan at any given moment.

Images