A bearing wear state monitoring-based life prediction method and system

Abstract

The application discloses a kind of life prediction method and system based on bearing wear state monitoring, comprising the following steps: step S1, obtain the wear state historical time series data of a plurality of bearings representing the entire operating life, and fit out life curve function based on the wear state historical time series data;Step S2, the wear state data of the target bearing to be predicted is input into the life curve function to predict the remaining life of the target bearing, the application uses the wear state historical time series data of a plurality of bearings representing the entire operating life to fit out life curve function, can realize only the wear state data of bearing is obtained to obtain the remaining life of bearing, and based on principal component extraction when the wear state historical time series data is acquired, only the state monitoring data with strong correlation with life is retained, realize the effective compression of data, guarantee the prediction correctness of life curve function while improving the fitting efficiency.

Description

Technical Field

[0001] This invention relates to the field of bearing testing technology for forklift masts, specifically to a life prediction method and system based on bearing wear condition monitoring. Background Technology

[0002] Forklift mast bearings are crucial hardware components used in forklifts, playing an irreplaceable role in their operation. Their remaining service life significantly impacts the overall lifespan of the forklift machinery. Therefore, condition monitoring and remaining service life estimation of bearings in critical components are essential. This requires understanding the degradation process and mechanism of forklift bearings, clarifying their damage evolution, providing a theoretical basis for developing suitable life prediction methods, and effectively improving the accuracy of life estimation.

[0003] To fully understand the instantaneous vibration behavior of forklift mast bearings caused by failure, lumped parameter models have been widely used. These models can simulate various types of bearing defects, such as surface roughness, surface waviness, dents, and spalling. However, current research focuses on the behavior of single failures in the short term and does not track the long-term degradation process of forklift mast bearings. Furthermore, it fails to effectively combine degradation mechanism characteristics with prediction methods, and the established models cannot be applied to life prediction. Summary of the Invention

[0004] The purpose of this invention is to provide a life prediction method and system based on bearing wear condition monitoring, in order to solve the technical problem that the existing technology does not effectively combine the characteristics of degradation mechanism with the prediction method, and the established model cannot be applied to life prediction.

[0005] To solve the above-mentioned technical problems, the present invention specifically provides the following technical solution:

[0006] A life prediction method based on bearing wear condition monitoring includes the following steps:

[0007] Step S1: Obtain historical time-series data of the wear and tear status of bearings of multiple forklift masts throughout their entire operating life, and fit a life curve function based on the historical time-series data of the wear and tear status.

[0008] Step S2: Input the wear status data of the target bearing to be predicted into the life curve function to predict the remaining life of the target bearing, and set a life warning factor for the target bearing to realize life warning when the remaining life of the target bearing is lower than the minimum warning line to ensure the safety of the target bearing in use.

[0009] In a preferred embodiment of the present invention, the fitting method for the lifetime curve function in step S1 includes:

[0010] Representative screening is performed on the wear status data at each historical time point in the wear status historical time series data of the multiple bearings to obtain standard wear status data at each historical time point. The standard wear status data represents that multiple bearings are located at the same historical time point and exhibit a consistent wear status.

[0011] The standard wear state data at each historical time series is linked temporally to obtain the standard wear state historical time series data, and the standard wear state historical time series data is input into a convolutional neural network for convolution fitting to obtain the lifetime curve function.

[0012] As a preferred embodiment of the present invention, the representative screening method includes:

[0013] Step S101: Position bearing j (j∈[1,m]) in the historical time series t i Loss status data at (i∈[1,n]) Each is individually quantized as located in the historical time series t i The dataset at (i∈[1,n]) The number of data sets is used as reference item A for merging;

[0014] Step S102: Set the merging threshold and merge the two datasets sequentially. and Similarity is calculated to obtain the similarity score, where,

[0015] If the similarity is higher than the merging threshold, then the dataset will be... and The data was merged to obtain a new dataset.

[0016] If the similarity is below the merging threshold, then there is no need to merge the dataset. and To merge;

[0017] Step S103: The quantity of the statistical data set is used as the combined control item B, where,

[0018] If B > w*A, then continue the merging iteration and return to step S102;

[0019] If B ≤ w * A, then output the current dataset as the result of the dataset merging.

[0020] Where i represents the historical time series identification number, j and k represent the bearing identification numbers, m represents the total number of bearings, and n represents the total number of time series. Characterized by bearing j being located in historical time series t i Data on the wear and tear status at the location, Characterized by bearing k being located in historical time series ti The data on the loss status at the location is represented by w, which is the weight of the control ratio.

[0021] As a preferred embodiment of the present invention, the representative screening method further includes:

[0022] The amount of data contained in each dataset in the merged dataset is counted, and the dataset with the largest amount of data is taken as the representative dataset.

[0023] The historical time series t is obtained by averaging all the wear and tear data in the representative dataset. i Standard loss state data at (i∈[1,n]) and the standard wear and tear data Linking the data yields historical time-series data of standard wear and tear conditions. in, Characterized as historical time series t i Standard wear and tear data at the location.

[0024] As a preferred embodiment of the present invention, the method for calculating the similarity of the dataset includes:

[0025] The average value of all the depleted state data in the dataset is used to obtain the dataset centroid, and the Euclidean distance is used to measure the distance between the centroids of the two datasets as the similarity between the two datasets.

[0026] The similarity calculation formula is as follows:

[0027]

[0028] In the formula, I op Represented as dataset C o and C p similarity, Represented as dataset C o and C p The loss status data at the center point, o and p (p≠o) are represented as the distinguishing numbers of the dataset.

[0029] As a preferred embodiment of the present invention, the convolution fitting method includes:

[0030] The historical time-series data of the standard wear and tear status Convert to time series sample form Among them, S ti Characterized as historical time series t i Standard wear and tear data at the location Characterized as historical time series t i The remaining lifespan at that location, t nCharacterized as the historical timeline of the end point of the runtime lifecycle;

[0031] The time series samples In As input to a convolutional neural network, As the output of the convolutional neural network, the convolutional neural network is constructed as a lifetime curve function ΔT=F(S,b), where F represents the nonlinear mapping function body of the worn-out state data S and the remaining lifetime ΔT, S represents the worn-out state data and serves as the function input, b represents the fluctuation constant of the function, and ΔT represents the remaining lifetime and serves as the function output.

[0032] In a preferred embodiment of the present invention, the method for predicting the remaining life of the target bearing in step S2 includes:

[0033] The remaining life of the target bearing is obtained by inputting the wear status data of the target bearing into the life curve function ΔT=F(S,b).

[0034] In a preferred embodiment of the present invention, the method for setting the lifespan warning factor in step S2 includes:

[0035] A lifespan warning factor is set, and the remaining lifespan of the target bearing is compared with the lifespan warning factor, wherein...

[0036] If the remaining lifespan is higher than the lifespan warning factor, there is no need to issue a lifespan warning for the target bearing.

[0037] If the remaining lifespan is lower than the lifespan warning factor, a lifespan warning will be issued for the target bearing.

[0038] As a preferred embodiment of the present invention, the method for obtaining the wear status data includes:

[0039] Acquire various condition monitoring data of the bearing, and perform principal component extraction on the various condition monitoring data to retain the condition monitoring data that is strongly correlated with the life as wear condition data;

[0040] The method for principal component extraction includes:

[0041] The various state monitoring data are fitted in two dimensions using data value-time series to obtain a set of state monitoring data curves. State monitoring data curves that show gentle fluctuations are then removed from the set of state monitoring data curves.

[0042] The remaining condition monitoring data curves in a set of condition monitoring data curves are used as the wear and tear condition data.

[0043] As a preferred embodiment of the present invention, the present invention provides a prediction system based on the aforementioned bearing wear condition monitoring-based life prediction method, comprising:

[0044] The curve fitting unit acquires historical time-series data of wear status of multiple bearings throughout their entire operating life, and fits a life curve function based on the historical time-series data of wear status.

[0045] The life prediction unit inputs the wear status data of the target bearing to be predicted into the life curve function to predict the remaining life of the target bearing.

[0046] The safety early warning unit sets a life warning factor for the target bearing to ensure the safety of its use by issuing a life warning when the remaining life of the target bearing is lower than the minimum warning line.

[0047] Compared with the prior art, the present invention has the following advantages:

[0048] This invention utilizes historical time-series data of the wear and tear status of bearings throughout the entire operational lifespan of multiple forklift masts to fit a lifespan curve function. This allows for the determination of the remaining lifespan of a bearing simply by acquiring its wear and tear status data. Furthermore, during the acquisition of the historical time-series data, principal component extraction is used to retain only state monitoring data strongly correlated with lifespan, achieving effective data compression. This ensures the accuracy of the lifespan curve function prediction while improving fitting efficiency. An early warning mode is also included to provide early warnings of bearing wear, enabling intelligent personnel to promptly replace and maintain the bearings, thus ensuring the safe operation of the forklift mast bearings. Attached Figure Description

[0049] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0050] Figure 1 A flowchart of the lifetime prediction method provided in an embodiment of the present invention;

[0051] Figure 2 This is a block diagram of the prediction system structure provided in an embodiment of the present invention.

[0052] The labels in the diagram represent the following:

[0053] 1-Curve fitting unit; 2-Lifetime prediction unit; 3-Safety early warning unit. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and 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.

[0055] like Figure 1 As shown, for a regular automated forklift mast structure, the bearings of forklift masts under the same working conditions in the same forklift mast structure will have a regular operating life. That is to say, the bearings of forklift masts under the same working conditions have the same outer ring wear state at various time sequences of their operating life. Therefore, this invention provides a life prediction method based on bearing wear state monitoring, constructs a life prediction function for the bearings of the forklift mast, and predicts the life of the bearings of the forklift mast.

[0056] A life prediction method based on bearing wear condition monitoring includes the following steps:

[0057] Step S1: Obtain historical time-series data of the wear and tear status of bearings of multiple forklift masts throughout their entire operating life, and fit a life curve function based on the historical time-series data of the wear and tear status.

[0058] The wear condition of a bearing determines its remaining life. Therefore, by establishing a mapping relationship between the wear condition and the remaining life, a life curve function can be fitted. By inputting the wear condition data of the bearing into the life curve function, the remaining life can be obtained.

[0059] To establish a mapping relationship between wear and remaining service life and to ensure the accuracy of this mapping, it is necessary to collect a large number of bearings that have completed their entire operational lifespan to obtain damage state data that is valuable for fitting the data. The specific collection method is as follows:

[0060] A large number of bearings were selected for operational testing, and a data acquisition device was set up for each bearing to collect various wear states in real time, obtaining multiple state monitoring data. For the bearings of the forklift mast, the research object of this implementation, the wear state acquisition operation involves rotating the outer ring of the forklift mast bearing and recording various state quantities of the forklift mast bearing during the outer ring rotation to obtain multiple state monitoring data. This process continues until the end of the entire operating life, obtaining historical time-series data of wear states that characterizes the operating life of the forklift mast bearing. The completeness of the large number of historical time-series data of wear states is screened, and historical time-series data of wear states that do not cover the entire operating life are removed, so that the retained historical time-series data of wear states have fitting value for the remaining life.

[0061] Condition monitoring data characterizes multiple wear states of bearings (such as outer ring surface roughness, outer ring surface waviness, outer ring indentation, outer ring spalling, outer ring rotation angle, etc.). Not all condition monitoring data affects bearing life; therefore, principal component analysis is required to extract the wear state data, identifying those with a high impact on bearing life. Methods for obtaining wear state data include:

[0062] Acquire various condition monitoring data of the bearing, and perform principal component extraction on the various condition monitoring data to retain the condition monitoring data that is strongly correlated with the life as wear condition data;

[0063] Methods for principal component extraction include:

[0064] A set of condition monitoring data curves is obtained by using data value-time series two-dimensional fitting of multiple condition monitoring data. The condition monitoring data curves that show gentle fluctuations are then removed from the set of condition monitoring data curves.

[0065] The remaining condition monitoring data curves in a set of condition monitoring data curves are used as the wear and tear condition data.

[0066] The above steps maintain the maximum range of types when collecting condition monitoring data. In order to find more types of condition monitoring data that may affect the lifespan, the scope of condition characteristic research is enriched. At the same time, the high-impact condition quantities are compressed and the low-impact condition monitoring data types are eliminated. Although this causes a slight decrease in the accuracy of the remaining life study, it effectively improves the efficiency of the remaining life study. When it is necessary to predict the bearing life during operation, it greatly improves the availability and safety.

[0067] In step S1, the fitting method for the lifetime curve function includes:

[0068] Representative screening of the wear status data at each historical time series in the wear status historical time series of multiple bearings is performed to obtain standard wear status data at each historical time series. The standard wear status data represents that multiple bearings at the same historical time series exhibit a consistent wear status.

[0069] The standard wear state data at each historical time series is linked temporally to obtain the standard wear state historical time series data, and the standard wear state historical time series data is input into a convolutional neural network for convolution fitting to obtain the lifetime curve function.

[0070] For multiple bearings that have completed their entire lifespan, the individual differences among bearings will lead to variations in the historical time-series data of their wear states. Since the fitted lifespan curve function needs to have universality, it is necessary to search for common features within the historical time-series data of multiple bearings. This allows for the search of representative standard wear state historical time-series data. Specifically, the wear state data at each historical time point in the standard wear state historical time-series data represents the standard value of the wear state data at each historical time point across multiple bearings. This standard wear state data characterizes the common wear state that the bearing actually possesses at that historical time point; most bearings will generate standard wear state historical data at that time point, making it a universally applicable data capture method.

[0071] Representative screening methods include:

[0072] Step S101: Position bearing j (j∈[1,m]) in the historical time series t i Loss status data at (i∈[1,n]) Each is individually quantized as located in the historical time series t i The dataset at (i∈[1,n]) The number of data sets is used as reference item A for merging;

[0073] Step S102: Set the merging threshold and merge the two datasets sequentially. and Similarity is calculated to obtain the similarity score, where,

[0074] If the similarity is higher than the merging threshold, then the dataset will be merged. and The data was merged to obtain a new dataset.

[0075] If the similarity is below the merging threshold, there is no need to merge the dataset. and To merge;

[0076] Step S103: The quantity of the statistical data set is used as the combined control item B, where,

[0077] If B > w*A, then continue the merging iteration and return to step S102;

[0078] If B ≤ w * A, then output the current dataset as the result of the dataset merging.

[0079] Where i represents the historical time series identification number, j and k represent the bearing identification numbers, m represents the total number of bearings, and n represents the total number of time series. Characterized by bearing j being located in historical time series t i Data on the wear and tear status at the location, Characterized by bearing k being located in historical time series t i The data on the consumption status at the point is represented by w, which is the weight of the control ratio. w is a percentage, such as 20%, which means that if the combined control term B is compared with 20% of the combined reference term A.

[0080] Representative screening methods also include:

[0081] The amount of data contained in each dataset in the merged dataset is counted, and the dataset with the largest amount of data is taken as the representative dataset.

[0082] The historical time series t is obtained by averaging all the wear and tear data in the representative dataset. i Standard loss state data at (i∈[1,n]) and standard wear and tear data Linking the data yields historical time-series data of standard wear and tear conditions. in, Characterized as historical time series t i Standard wear and tear data at the location.

[0083] Methods for calculating the similarity of datasets include:

[0084] The average value of all the depleted state data in the dataset is used to obtain the dataset centroid, and the Euclidean distance is used to measure the distance between the centroids of the two datasets as the similarity between the two datasets.

[0085] The formula for calculating similarity is:

[0086]

[0087] In the formula, I op Represented as dataset C o and C p similarity, Represented as dataset C o and C p The loss status data at the center point, o and p (p≠o) are represented as the distinguishing numbers of the dataset.

[0088] For example, the wear status data of bearings 1, 2, and 3 at historical time t1. Quantize the wear status data into a set The midpoints of C1 and C2 are calculated as follows: The similarity between C1 and C2 is If the similarity between C1 and C2 is higher than the merging threshold, then C1 and C2 will be merged to obtain a new set. Calculate the new set midpoint The similarity between the new set C1+C2 and set C3 is calculated. If the similarity is below the merging threshold, then the original set format is maintained. Then set For a representative set, then This is standard wear and tear data.

[0089] Convolution fitting methods include:

[0090] Historical time-series data of standard wear and tear status Convert to time series sample form in, Characterized as historical time series t i Standard wear and tear data at the location Characterized as historical time series t i The remaining lifespan at that location, t n Characterized as the historical timeline of the end point of the runtime lifecycle;

[0091] Time series samples S in ti As input to a convolutional neural network, As the output of the convolutional neural network, the convolutional neural network is constructed as a lifetime curve function ΔT=F(S,b), where F represents the nonlinear mapping function body of the worn-out state data S and the remaining lifetime ΔT, S represents the worn-out state data and serves as the function input, b represents the fluctuation constant of the function, and ΔT represents the remaining lifetime and serves as the function output.

[0092] Convolutional neural networks can fit the mapping relationship when there is sufficient sample data. Therefore, by inputting a large number of standard wear-out state historical time series data and remaining lifetime into the convolutional upgrade network for convolutional fitting, a nonlinear mapping relationship between the standard wear-out state historical time series data and the remaining lifetime can be obtained, thereby realizing the function of calculating the remaining lifetime by inputting wear-out state data.

[0093] In step S2, the method for predicting the remaining life of the target bearing includes:

[0094] The remaining life of the target bearing is obtained by inputting the wear status data of the target bearing into the life curve function ΔT=F(S,b).

[0095] Step S2: Input the wear status data of the target bearing to be predicted into the life curve function to predict the remaining life of the target bearing, and set a life warning factor for the target bearing to realize life warning when the remaining life of the target bearing is lower than the minimum warning line to ensure the safety of the target bearing in use.

[0096] In step S2, the method for setting the lifespan warning factor includes:

[0097] A lifespan warning factor is set, and the remaining lifespan of the target bearing is compared with the lifespan warning factor.

[0098] If the remaining lifespan is higher than the lifespan warning factor, there is no need to issue a lifespan warning for the target bearing.

[0099] If the remaining lifespan is lower than the lifespan warning factor, a lifespan warning will be issued for the target bearing.

[0100] This embodiment does not limit the lifespan warning factor; it can be customized by the user.

[0101] like Figure 2 As shown, based on the above-mentioned life prediction method based on bearing wear condition monitoring, the present invention provides a prediction system, including:

[0102] Curve fitting unit 1 acquires historical time-series data of wear status of multiple bearings throughout their entire operating life, and fits a life curve function based on the historical time-series data of wear status.

[0103] The life prediction unit 2 inputs the wear status data of the target bearing to be predicted into the life curve function to predict the remaining life of the target bearing.

[0104] Safety warning unit 3 sets a life warning factor for the target bearing to ensure the safety of the target bearing in use by issuing a life warning when the remaining life of the target bearing is lower than the minimum warning line.

[0105] This invention utilizes historical time-series data of the wear and tear status of bearings throughout the entire operational lifespan of multiple forklift masts to fit a lifespan curve function. This allows for the determination of the remaining lifespan of a bearing simply by acquiring its wear and tear status data. Furthermore, during the acquisition of the historical time-series data, principal component extraction is used to retain only state monitoring data strongly correlated with lifespan, achieving effective data compression. This ensures the accuracy of the lifespan curve function prediction while improving fitting efficiency. An early warning mode is also included to provide early warnings of bearing wear, enabling intelligent personnel to promptly replace and maintain the bearings, thus ensuring the safe operation of the forklift mast bearings.

[0106] The above embodiments are merely exemplary embodiments of this application and are not intended to limit this application. The scope of protection of this application is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to this application within its substance and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of this application.

Claims

1. A life prediction method based on bearing wear condition monitoring, characterized in that: Includes the following steps: Step S1: Obtain historical time-series data of the wear and tear status of bearings of multiple forklift masts throughout their entire operating life, and fit a life curve function based on the historical time-series data of the wear and tear status. Step S2: Input the wear status data of the target bearing to be predicted into the life curve function to predict the remaining life of the target bearing, and set a life warning factor for the target bearing to realize life warning when the remaining life of the target bearing is lower than the minimum warning line to ensure the safety of the target bearing in use. In step S1, the fitting method for the lifetime curve function includes: Representative screening is performed on the wear status data at each historical time point in the wear status historical time series data of the multiple bearings to obtain standard wear status data at each historical time point. The standard wear status data represents that multiple bearings are located at the same historical time point and exhibit a consistent wear status. The standard wear state data at each historical time series is linked in time series to obtain the standard wear state historical time series data, and the standard wear state historical time series data is input into a convolutional neural network for convolution fitting to obtain the lifetime curve function. The representative screening method includes: Step S101: Place the bearing , Located in historical timeline , Data on wear and tear at the location Each one is individually quantified as being located in the historical time series. , Data set at the location The number of data sets is used as the merging reference item A; Step S102: Set the merging threshold and merge the two datasets sequentially. and , Similarity is calculated to obtain the similarity score, where, If the similarity is higher than the merging threshold, then the dataset will be... and , The data was merged to obtain a new dataset. ; If the similarity is below the merging threshold, then there is no need to merge the dataset. and , To merge; Step S103: The quantity of the statistical data set is used as the combined control item B, where, like If so, continue the merging iteration and return to step S102; like If the current dataset is not found, then the current dataset will be output as the merged dataset result. in, The representation is based on historical time sequence and numbering. , The designation is based on the bearing identification number. Characterized by the total number of bearings, Characterized by the total number of time series, Characterized as bearings Located in historical timeline Data on the wear and tear status at the location, Characterized as bearings Located in historical timeline Data on the wear and tear status at the location, This is represented by the adjustment ratio weight.

2. The life prediction method based on bearing wear condition monitoring according to claim 1, characterized in that: The representative screening method also includes: The amount of data contained in each dataset in the merged dataset is counted, and the dataset with the largest amount of data is taken as the representative dataset. The historical time series is obtained by averaging all the wear and tear data in the representative dataset. , Standard wear and tear data and the standard wear and tear status data Linking the data yields historical time-series data of standard wear and tear conditions. ,in, Represented as historical time sequence Standard wear and tear data at the location.

3. The life prediction method based on bearing wear condition monitoring according to claim 1, characterized in that: The similarity calculation method for the dataset includes: The average value of all the depleted state data in the dataset is used to obtain the dataset centroid, and the Euclidean distance is used to measure the distance between the centroids of the two datasets as the similarity between the two datasets. The similarity calculation formula is as follows: ； In the formula, Represented as a dataset and similarity, Represented as a dataset and The loss status data at the center point, o and p, It is represented by the distinguishing number of the dataset.

4. The life prediction method based on bearing wear condition monitoring according to claim 1, characterized in that: The convolution fitting method includes: The historical time-series data of the standard wear and tear status Convert to time series sample form ,in, Represented as historical time sequence Standard wear and tear data at the location Represented as historical time sequence The remaining lifespan at that location, , Characterized as the historical timeline of the end point of the runtime lifecycle; The time series samples In As input to a convolutional neural network, As the output of the convolutional neural network, the convolutional neural network is constructed as a lifetime curve function. ,in, Characterized as depletion state data and remaining lifespan The body of the nonlinear mapping function, The data is represented as loss state data and used as the input term of the function, where b represents the fluctuation constant of the function. It is represented as the remaining lifetime and is used as the output term of the function.

5. The life prediction method based on bearing wear condition monitoring according to claim 4, characterized in that: In step S2, the method for predicting the remaining life of the target bearing includes: The wear status data of the target bearing is used as input to input the life curve function. The remaining life of the target bearing is obtained.

6. The life prediction method based on bearing wear condition monitoring according to claim 1, characterized in that: In step S2, the method for setting the lifespan warning factor includes: A lifespan warning factor is set, and the remaining lifespan of the target bearing is compared with the lifespan warning factor, wherein... If the remaining lifespan is higher than the lifespan warning factor, there is no need to issue a lifespan warning for the target bearing. If the remaining lifespan is lower than the lifespan warning factor, a lifespan warning will be issued for the target bearing.

7. The life prediction method based on bearing wear condition monitoring according to claim 1, characterized in that, The method for obtaining the wear status data includes: acquiring multiple condition monitoring data of the bearing, and performing principal component extraction on the multiple condition monitoring data to retain the condition monitoring data that is strongly correlated with the life as the wear status data; The method for principal component extraction includes: The various state monitoring data are fitted in two dimensions using data value-time series to obtain a set of state monitoring data curves. State monitoring data curves that show gentle fluctuations are then removed from the set of state monitoring data curves. The remaining condition monitoring data curves in a set of condition monitoring data curves are used as the wear and tear condition data.

8. A prediction system for a life prediction method based on bearing wear condition monitoring according to any one of claims 1-7, characterized in that, include: Curve fitting unit (1) acquires historical time-series data of wear status of multiple bearings throughout their entire operating life, and fits a life curve function based on the historical time-series data of wear status. The life prediction unit (2) inputs the wear status data of the target bearing to be predicted into the life curve function to predict the remaining life of the target bearing. The safety warning unit (3) sets a life warning factor for the target bearing to ensure the safety of the target bearing when the remaining life of the target bearing is lower than the minimum warning line.

Images