System, device and method for determining bearing condition

Abstract

Systems, apparatuses, and methods of determining a condition of at least one bearing (810) in a system are disclosed. The method includes receiving operational data associated with a system from one or more sensing units associated with the system; determining an operational profile of the at least one bearing (810) from the operational data, wherein the operational profile includes at least one of a vibration response, a thermal response, and a frequency response associated with the at least one bearing (810); determining an impact force profile during operation of the at least one bearing (810) based on the operational profile and a virtual bearing model (400, 600), the virtual bearing model trained on operational profiles and impact force profiles associated with a group of bearings comparable to the at least one bearing (810); and determining a condition of the at least one bearing (810) based on the impact force profile.

Description

[0001] The present invention relates to determining the condition of bearings in a system.

[0002] Bearings used in motors, or any rotating system, can fail for a variety of reasons. For example, in the case of rolling element bearings, their service life can be affected by a loss of lubricant function. Lubricants can lose their lubricating ability due to concentrated stresses that can cause temperature increases. Lubricants can also lose their lubricating ability due to lubricant contamination caused by wear-generated particles.

[0003] Techniques for determining lubricant condition include experimental methods. These methods are based on data related to the operation of the bearing or the system in which the bearing is used. The accuracy of the data-based analysis depends on the placement of the sensing elements capturing the data. For example, the vibration response of a bearing can depend on the placement of an accelerometer.

[0004] Furthermore, in some scenarios, it's impossible to measure the stresses that could lead to a loss of lubrication. In such scenarios, data-based analysis may not accurately indicate the condition of the lubricant. Consequently, the bearing's condition may not be accurately identified.

[0005] In view of the above, there exists a need to determine the condition of a bearing. Therefore, an object of the present invention is to provide a system, apparatus and method for determining the condition of a bearing in a system.

[0006] An aspect of the present invention is a computer-implemented method for determining a condition of at least one bearing in a system. The method includes receiving operational data associated with the system from one or more sensing units associated with the system; determining an operational profile of the at least one bearing from the operational data, wherein the operational profile includes at least one of a vibration response, a thermal response, and a frequency response associated with the at least one bearing; determining an impact force profile during operation of the at least one bearing based on the operational profile and a virtual bearing model, the virtual bearing model being trained on operational profiles and impact force profiles associated with a group of bearings comparable to the at least one bearing; and determining the condition of the at least one bearing based on the impact force profile.

[0007] Example bearings include fluid bearings and rolling element bearings with rolling elements or needle rollers. Example systems include rotors, motors, transmissions, gearboxes, etc.

[0008] As used herein, "operational data" refers to data received from various sources (e.g., sensors, scanners, user devices, etc.) that reflects the operating condition of a bearing and / or system. Sensors measure operating parameters associated with a technical system. Sensors may include vibration sensors, current and voltage sensors, etc. For example, the measurement of shaft voltage in a motor is mapped to an operating parameter of a bearing. The term "operational parameter" refers to one or more characteristics of a bearing. For example, operational data may include values ​​of vibration, temperature, current, magnetic flux, speed, and power of the system (including the bearing).

[0009] As used herein, an "operational profile" refers to a combination of one of a vibration response, a thermal response, a frequency response, a magnetic response, etc. The response is in turn generated from the operational data. In an embodiment, the vibration response may be generated based on the root mean square of the vibration data from the vibration sensor.

[0010] As used herein, "impact force profile" refers to the impact force measured and determined based on the operating profile. The impact force is measured during the operation of the bearing and is therefore referred to as the impact force profile.

[0011] As used herein, a "virtual bearing model" refers to a software-defined bearing generated based on operational data of a bearing set and physical properties associated with the bearing. The virtual bearing model includes predictive and artificial intelligence algorithms to predict the condition of the bearing.

[0012] The method may include predicting a stress distribution associated with the at least one bearing during operation. The stress distribution is predicted based on an impact force profile of the at least one bearing and a virtual bearing model. Furthermore, the method may include predicting a remaining life of the at least one bearing based on the stress distribution and the predicted life using a neural network; wherein the neural network is configured to perform gradient descent optimization.

[0013] The method may include identifying defects in the at least one bearing based on the impact force profile and determining a contamination condition of lubricant in the at least one bearing. In an embodiment, the defects are identified and the contamination condition is determined by superimposing an operating profile of the at least one bearing on an operating profile in a virtual bearing model. Furthermore, the superimposition is performed by deriving an impact force profile and determining a stress distribution.

[0014] The method may further include determining fatigue of the at least one bearing based on the virtual bearing model with respect to at least one of a lubricant temperature increase, foreign particles in the lubricant, and a decrease in an oil film parameter of the lubricant.

[0015] The second aspect of the present disclosure includes a computer-implemented method of generating a virtual bearing model. The method includes determining a test operating profile based on test operating data associated with a bearing set; simulating a predetermined defect on a predefined bearing model, the predefined bearing model comprising a data set in accordance with dynamic load rating criteria and rated life criteria associated with the bearing set; and generating a simulated operating profile associated with the bearing set based on the simulation of the predetermined defect on the predefined bearing model; wherein the test operating profile and the simulated operating profile comprise a vibration response, a thermal response, and a frequency response associated with the bearing set.

[0016] As used herein, "test operating data" refers to operating data generated from a bearing set during a bearing test. The test operating data includes vibration, temperature, current, magnetic flux, speed, and power values of the system, including the bearing. The "test operating data" is different from "operating data" based on a source bearing.

[0017] As used herein, "predefined bearing model" refers to a model generated based on dynamic load rating criteria and rated life criteria associated with a bearing set. For example, the predefined bearing model is a physics-based model generated through finite element modeling.

[0018] The virtual bearing model can include a life prediction algorithm. Accordingly, the method can include predicting a life of a bearing when subjected to the predetermined defect based on at least one of a bearing load, a load zone, a bearing clearance, a lubrication viscosity, and a lubricant contamination associated with one or more bearings in the bearing set.

[0019] In an embodiment, to generate the virtual bearing model, the method can include implementing one or more predetermined defects on one or more bearings in the bearing set. The predetermined defects include a lubricant contamination, a Brinell indentation on a bearing raceway, or a flaking damage on the raceway. The predetermined defects can be accurately created using techniques such as electrical discharge machining (EDM) and laser engraving.

[0020] Further, the method can include operating a system including the bearing under one or more system load conditions. The system load conditions indicate a system load on the system. For example, the system load conditions include a case of presence or absence of a load.

[0021] Still further, the method can include determining the test operating profile associated with the bearing set for the system load conditions, the test operating profile being generated based on test operating data received from a radial positioning, an axial positioning, and a horizontal positioning.

[0022] The method can comprise determining a simulated impact force from the test operating profile and the simulated operating profile. Further, the method can comprise predicting the simulated impact force based on the simulated operating profile and at least one mass of the bearing rolling elements, a damping coefficient and a stiffness associated with the bearing. The simulated impact force comprises at least one of a steady state component from a steady state movement of the bearing and a dynamic component associated with the predetermined defect.

[0023] In an embodiment, the method comprises predicting the simulated impact force based on the simulated operating profile and at least one mass of the bearing rolling elements, a damping coefficient and a stiffness associated with the bearing, wherein the simulated impact force comprises at least one of a steady state component from a steady state rotation and a dynamic component associated with the predetermined defect.

[0024] The method can comprise predicting a stress profile associated with the set of bearings based on a comparison of the test operating profile and the simulated operating profile to update the stress profile. In an embodiment, the stress profile is predicted by performing the following steps, namely comparing the test operating profile and the simulated operating profile; updating the simulated impact force based on the comparison; generating the stress profile based on the updated simulated impact force; and mapping the stress profile to the predetermined defect using at least one machine learning algorithm.

[0025] To update the simulated impact force, the method comprises calibrating the updated simulated impact force associated with the set of bearings using at least one machine learning algorithm based on a difference between the test operating profile and the simulated operating profile.

[0026] In an embodiment, the method can comprise calibrating the updated simulated impact force using a differential evolution algorithm. Accordingly, the method can further comprise defining an upper stress limit and a lower stress limit of the set of bearings; determining a possible stress profile within the upper stress limit and the lower stress limit by one of a mutation and recombination operation, wherein the possible stress profile is determined for the difference; and selecting the stress profile from the possible stress profile using a continuous function based on the impact force difference.

[0027] The method facilitates generating a virtual bearing model with an accurate remaining life prediction in case of contamination and loss of lubrication. The above described method is a combination of simulation based on physical properties and machine learning methods. The method of superimposing the operating profile of at least one bearing and the operating profile of the virtual bearing model enables an accurate estimation of the remaining life of the at least one bearing. Further, the knowledge generated from the set of bearings is used to generate the virtual bearing model. This enables the knowledge of the fleet to be used for calibrating the remaining life. The above described method can be used to guarantee the availability of bearings in a system.

[0028] A third aspect of the present invention comprises an apparatus for determining a condition of at least one bearing in a system. The apparatus comprises one or more processing units; and a memory unit communicatively coupled to the one or more processing units. The memory unit comprises a bearing module and a virtual bearing module, the bearing module and virtual bearing module being stored in the form of machine readable instructions executable by the one or more processing units. Furthermore, the bearing module is configured to perform one or more method steps associated with the at least one bearing, and the virtual bearing module is configured to perform a method of generating a virtual bearing model.

[0029] A fourth aspect of the present invention comprises a system comprising one or more devices capable of providing operational data associated with the operation of one or more systems; and an apparatus communicatively coupled to the one or more devices, wherein the apparatus is configured for determining a condition of at least one bearing in the one or more systems.

[0030] A fifth aspect of the present invention comprises a computer program product having machine readable instructions stored therein, the instructions, when executed by one or more processing units, cause the one or more processing units to perform the above method.

[0031] The above and other features of the present invention will be described now with reference to the drawings of the present invention. The illustrated embodiments are intended to illustrate, not limit, the present invention.

[0032] The present invention will be described further with reference to the embodiments shown in the drawings, wherein:

[0033] Figure 1 illustrates the stages associated with the deterioration of a bearing in a system according to embodiments of the present invention;

[0034] Figure 2 illustrates the relationship between the remaining life of a bearing relative to the condition of a lubricant according to embodiments of the present invention;

[0035] Figure 3 illustrates a method 300 of generating a virtual bearing model according to embodiments of the present invention;

[0036] Figure 4 illustrates a virtual bearing model 400 of a ball in a rolling bearing according to embodiments of the present invention;

[0037] Figure 5 illustrates the stages in a shock cycle of a rolling bearing in Figure 4

[0038] Figure 6 illustrates a virtual bearing model 600 for a rolling bearing having a plurality of balls according to embodiments of the present invention;

[0039] illustrates a virtual bearing model 600 for a rolling bearing having a plurality of balls according to embodiments of the present invention; Figure 7 FIG. 7 illustrates a method 700 of determining a condition of a bearing in a system, in accordance with an embodiment of the application;

[0040] Figure 8 FIG. 8 illustrates an apparatus 820 for determining a condition of a bearing 810 in a system 800 at runtime, in accordance with an embodiment of the application; and

[0041] Figure 9 FIG. 9 illustrates a system 900 for determining a condition of a plurality of bearings 912, 922, and 932 in one or more systems 910, 920, and 930, in accordance with an embodiment of the application.

[0042] In the following, embodiments for carrying out the application will be described in detail. Various embodiments are described with reference to the accompanying drawings, wherein like reference numerals are used throughout to denote like elements. In the following description, for the purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of one or more embodiments. It can be apparent, however, that such embodiments can be practiced without

[0043] “Operational data” as used in the following refers to data reflecting the operational condition of a bearing and / or a system, which is received from different sources (e.g., sensors, scanners, user devices, etc.). Sensors measure operational parameters related to a technical system. Sensors can include vibration sensors, current and voltage sensors, etc. For example, a measurement of shaft voltage in a motor is mapped to an operational parameter of a bearing. For example, operational data includes values of vibration, temperature, current, magnetic flux, speed, power of a system (including a bearing).

[0044] “Virtual bearing model” as used in the following refers to a software-defined bearing generated based on operational data of a bearing set and physical characteristics associated with the bearing. The virtual bearing model includes predictive and artificial intelligence algorithms to predict a condition of a bearing.

[0045] “Remaining life” as used in the following refers to the life of a bearing with defects and contamination. Remaining life includes remaining useful life (RUL), downtime, maintenance time, etc. “Remaining life” is different from “life”. “Life” refers to the life of a bearing without defects or life at the beginning of use.

[0046] “Bearing condition” as used in the following refers to the state of a bearing. For example, a bearing condition includes the presence of defects in a bearing, a contamination condition, a remaining life, etc.

[0047] Figure 1The diagram illustrates stages 102-112 associated with bearing condition deterioration in a system according to an embodiment of the present invention. Stages 102-112 serve as a framework for accurately predicting the remaining life of a bearing. Stages 102-112 are determined based on operational data associated with the bearing and / or the system. Furthermore, stages 102-112 are determined based on a virtual bearing model of the bearings in the system.

[0048] Stage 102 indicates a "normal condition" of the bearing. In stage 102, the bearing is in good condition and the lubricant in the bearing is not contaminated. Stage 102 also includes a condition where the bearing is normally contaminated. The bearing condition in stage 102 can be determined based on operational data.

[0049] Stage 104 indicates "light contamination" of the lubricant. In stage 104, the bearing condition transitions from normal contamination to light contamination. The transition between normal contamination and light contamination can be difficult to estimate. Therefore, a combination of operating data and a virtual bearing model is analyzed to determine the transition.

[0050] Stage 106 indicates "heavy contamination" of the lubricant. The bearing condition changes from light contamination to heavy contamination in stage 106. Stage 106 is determined based on a combination of operating data and a virtual bearing model.

[0051] Stage 108 indicates the "temperature condition" of the lubricant. In stage 108, the lubricant is able to maintain a predetermined temperature for a certain duration. For example, if the lubricant loses its lubricating ability at 100°C, then the predetermined temperature is below 100°C.

[0052] Stage 110 indicates the "system status" of the system using the bearing. In stage 110, the status of the system is determined based on the operational data to determine any fault conditions. The virtual bearing model is used to determine whether the fault condition is associated with the bearing.

[0053] Stage 112 describes the "load conditions" of the system using the bearing. In stage 112, the system's conditions are determined for various load conditions. Based on the virtual bearing model, the load conditions are mapped to bearing parameters associated with the bearing. For example, the load conditions are mapped to the required speed change of the bearing.

[0054] At step 114 , analysis of the operational data and the virtual bearing model is performed. Figure 8 This analysis is further described in

[15] . Furthermore, at step 116, the remaining life of the bearing is predicted based on the analysis. At step 118, test operating data from a similar bearing is received. The test operating data is compared with the operating data and the virtual bearing model. This comparison is used at step 120 to calibrate the remaining life, resulting in an accurate remaining life prediction.

[0055] Figure 2A graph illustrates a relationship 210 between bearing life 220 versus lubricant condition. The life 220 is determined at the beginning of use of the bearing. The lubricant condition is indicated in 4 stages, namely normal contamination 212, mild contamination 214, severe contamination 216, and extreme contamination 218. The life 220 is calculated based on the number of cycles to failure. The relationship 210 is determined for multiple load conditions 4KN, 3kN, 2kN, and lkN.

[0056] For example, Figure 2 The remaining life 222 for normal contamination is illustrated as infinite. For mild contamination, the remaining life 224 is 42500 cycles to failure for 4KN load, and infinite for 3kN-lkN. For severe contamination, the remaining life 226 is 22600 cycles for 4kN, 66210 cycles for 3kN, and infinite for 2kN and lkN. For extreme contamination, the remaining life 228 is 8730 cycles for 4kN load, 22800 cycles for 3kN load, 89700 cycles for 2kN load, and indeterminate for lkN load.

[0057] As Figure 2 shown in the graph 210, the life of the bearing is infinite. The relationship 210 can not accurately predict the life 220. Therefore, test operating data is used to determine the remaining life 250 in normal use. The relationship 210 can be updated based on the remaining life 250. In this example, the remaining life 250 is estimated to be 40000 cycles to failure.

[0058] Figure 2 A table column 260 is also included that indicates the relationship 210. In addition, the relationship 210 can also be illustrated by a graph 270.

[0059] Figure 3 A method 300 of generating a virtual bearing model is illustrated in accordance with an embodiment of the present application. The virtual bearing model is a trained model that is generated from a group of bearings having comparable operating parameters. The group of bearings can each be housed in a bearing housing and can be provided in one or more systems. For example, the group of bearings can be rolling bearings provided in one or more rotating machines.

[0060] The term "operating parameter" refers to one or more characteristics of the bearing. For example, the operating parameters include vibration, temperature, current, magnetic flux, speed, power values of the system (including the bearing).

[0061] The method 300 includes generating a virtual bearing model based on both test-based modeling 305 and simulation-based modeling 308. Steps 302-306 relate to test-based modeling, and step 308 relates to simulation-based modeling. Those skilled in the art will appreciate that the techniques can be performed in parallel or sequentially without a substantial impact on the virtual bearing model generated.

[0062] At step 302, one or more predetermined defects on one or more bearings of a bearing set are affected. For the purposes of the following explanation, the predetermined defects affect each bearing of the bearing set. The predetermined defects include lubricant contamination, Brinell indentation on a bearing raceway, or spalling damage on a raceway. Those skilled in the art will appreciate that the predetermined defects can vary from bearing to bearing. The predetermined defects can be created accurately using techniques such as electrical discharge machining (EDM) and laser engraving.

[0063] At step 304, a system including the bearings is operated under one or more system load conditions. As used herein, "system load condition" refers to a system load on a system including the bearings. Further, the system load condition indicates whether the system is operating under a system load.

[0064] At step 306, a test operating profile associated with the bearing set is generated. The test operating profile is generated based on test operating data for the system load conditions. The test operating data is received from one or more sensing units located radially, axially, and horizontally external and internal to bearing housings associated with each of the bearings.

[0065] As used herein, the test operating profile refers to a vibration response, a thermal response, and / or a frequency response generated from the test operating data. In an embodiment, the vibration response from the bearing set is referred to as the test operating profile.

[0066] At step 308, the predetermined defects are simulated on a predefined bearing model. The predefined bearing model includes a data set that complies with dynamic load rating criteria and rated life criteria associated with the bearing set. The dynamic load rating criteria and rated life criteria are based on physical properties.

[0067] At step 308, further, a simulated operating profile associated with the bearing set is generated. The simulated operating profile includes a vibration response, a thermal response, and / or a frequency response. The simulated operating profile is generated in response to the simulation of the predetermined defects on the predefined model.

[0068] At step 310, a simulated impact force is predicted based on the simulated operating profile. The simulated impact force is also based on at least one mass of a bearing rolling element (such as a needle roller or a ball), a damping coefficient, and a stiffness associated with the bearing. The simulated impact force includes a stable component resulting from steady-state movement of the bearing. In addition, the simulated impact force includes a dynamic component associated with the impact caused by the predetermined defect. The simulated impact force is determined in Figure 6 It is explained in.

[0069] Furthermore, the test operation profile is compared to the simulated operation profile. The simulated impact force is updated based on the comparison. A machine learning algorithm, such as a genetic algorithm, is used to compare the test operation profile to the simulated operation profile. In an embodiment, a differential evolution algorithm is used to update the simulated operation profile.

[0070] At step 320 , a stress distribution is generated based on the simulated impact force. For example, Hertz contact stress theory is used to determine the stress distribution based on the impact force.

[0071] At step 330, the stress distribution is updated based on the impact force difference. In one embodiment, a machine learning algorithm, such as a differential evolution algorithm, is used to update the stress distribution. The differential evolution algorithm is used to determine the limits of the stress distribution. When determining the upper and lower stress limits, possible stress distributions within the upper and lower stress limits are determined through one of mutation and recombination operations.

[0072] A stress distribution is selected from the possible stress distributions using continuous function optimization based on impact force differences. The impact force differences are used to generate an optimization problem for the differential evolution algorithm. The optimization problem is used to narrow down the possible stress distributions based on fitness scores to the optimization problem.

[0073] At step 340, the updated stress distribution is mapped to the predetermined defect using a differential evolution algorithm. The operations performed include limit setting, mutation, recombination, and selection. The output of the above operations results in the stress distribution being mapped to the predetermined defect.

[0074] The life of the bearing when subjected to a predetermined defect is predicted at step 350. The life is predicted based on stress distribution, bearing load, load zone, bearing clearance, lubrication viscosity, and lubricant contamination associated with the bearing set.

[0075] Thus, the virtual bearing model generated in method 300 is configured to be used to determine the remaining life of the unknown bearing based on operational data associated with the unknown bearing. In addition, the virtual bearing model is capable of identifying impact forces and defects in the unknown bearing. A detailed description of the use of the virtual bearing model is provided in Figure 7 described in .

[0076] Figure 4A virtual bearing model 400 of a ball in a rolling element bearing according to an embodiment of the present invention is illustrated. The virtual bearing model 400 includes a simulated ball 402 on an outer raceway 404 having a simulated defect 450. The defect 450 includes a leading edge 406 and a trailing edge 408. The leading edge 406 is referred to as the entry edge, and the trailing edge 408 is referred to as the exit edge.

[0077] The virtual bearing model 400 is generated based on associated boundary conditions. For example, the boundary conditions may include securing the rolling bearing with a bolted joint to provide surface-to-surface contact with an appropriate friction coefficient and rotational frequency for the inner race.

[0078] The virtual bearing model 400 is used to determine the impact force of a rolling bearing on a physical defect. The impact force is determined based on the movement of a simulated ball 402 on a simulated defect 450. The detailed description of the movement is given in Figure 5 Middle picture.

[0079] Figure 5 Pictured Figure 4 4. The stages in an impact cycle 500 of a rolling element bearing on the leading edge 406 and the trailing edge 408. The impact cycle 500 includes two pulses 510 and 520 at the leading edge 406 and the trailing edge 408. The pulse 520 is generally higher than the pulse 510.

[0080] For example, a rolling bearing is in a rotor with a load condition of a static load of 400N. In addition, the shaft in the rotor is running at a speed of 1478 revolutions per minute. The virtual bearing model is used to determine the impact force of the rolling bearing based on the movement of the simulated ball 402 on the leading edge 406 and the trailing edge 408 on the simulated defect 450. Therefore, the leading edge force 412 is determined to be 4.07kN and the trailing edge force 414 is determined to be 5.39kN. The leading edge force 412 is observed at 0.17991ms and the trailing edge force is observed at 0.24993ms. The time difference between the leading edge force 412 and the trailing edge force 414 is used to determine the defect size. This is in Figure 6 Explained in.

[0081] Figure 6 A virtual bearing model 600 for a rolling bearing having a plurality of balls according to an embodiment of the present invention is illustrated. Figure 6 The virtual bearing model 600 in FIG. 5 illustrates a simulated ball 602 on an outer raceway 604. The virtual bearing model 600 also illustrates a simulated defect 650 on the outer raceway 604 having a leading edge 606 and a trailing edge 608.

[0082] The virtual bearing model 600 is configured to illustrate the impact forces generated due to simulated defects 650 at the leading edge 606 and the trailing edge 608. The impact forces are illustrated in graph 620. The leading edge force is referenced by numeral 622, and the trailing edge force is referenced by numeral 624.

[0083] Virtual bearing model 600 predicts the impact forces illustrated in graph 620. The impact force prediction assumes steady-state rotation. Most rolling bearing applications involve steady-state rotation on the outer and / or inner raceways. The rotational speed can be moderate to avoid centrifugal forces or significant gyroscopic motion of the balls. The magnitude of the impact felt by the rolling bearing when the balls pass through the defect depends on the relative speed and the applied external load.

[0084] Based on the above, the impact force should produce a static component and a dynamic component. The static component is developed using the following equation, and the dynamic component is produced by the impact force of the simulated ball 602 on the edges 606 and 608 of the simulated defect 650.

[0085]

[0086] otherwise,

[0087]

[0088] in

[0089] m is the mass of the ball

[0090] mg is the mass of the ball and the acceleration due to gravity

[0091] B d is the ball diameter

[0092] I is the mass moment of inertia

[0093] V is the linear velocity of the ball

[0094] is the angular velocity of the ball.

[0095] Furthermore, the impact force depends on the load zone of the rolling bearing. The virtual bearing model 600 takes into account whether the simulated ball 602 is in the load zone and whether the outer raceway 604 and the inner raceway ( Figure 6 Therefore, the impact force is determined based on the following equation.

[0096] Impact forces from experiments where an object freely falls onto a steel plate show that the impact force varies as the square of the impact velocity, which is used to determine the impact forces 622 and 624 .

[0097]

[0098] in

[0099] is a constant that depends on the impact material and the value of the falling mass

[0100]

[0101] in

[0102] is a constant that depends only on the impact material

[0103] F s is the static force from the falling mass

[0104]

[0105] Equations considering impact forces Q i , the equation for the total impact force can be determined as follows.

[0106]

[0107] in

[0108] is the impact coefficient, which depends on both the impact material and the bearing geometry.

[0109] The simulated defect 650 can have a variety of widths. The width can be calculated inversely based on the velocity of the simulated ball 602 and the time difference between the leading edge force 622 and the trailing edge force 624. Figure 6 , the defect width 625 of the simulated defect 650 is determined based on the velocity of the simulated ball 602 and the time difference.

[0110] In an embodiment, when the defect width 625 is known, the impact forces 622 and 624 can be determined as follows. In this embodiment, the impact forces 622 and 624 are considered to be proportional to the square of the defect width 625. Therefore, F T Export as follows.

[0111]

[0112] in

[0113] B d is the rolling element diameter

[0114] k is the proportionality constant

[0115] d def is the defect width.

[0116] Figure 7A method 700 of bearing condition in a system is illustrated according to an embodiment of the present invention. Figure 7 The bearings in Figure 3 Therefore, method 700 can be used when there is no historical data on the bearing condition.

[0117] Method 700 begins at step 702. At step 702, operational data associated with the system is received. The operational data is generated from one or more sensing units associated with the system. The operation of the system may reflect the condition of the bearing. Therefore, the operational data of the system is used to analyze the condition of the bearing.

[0118] At step 704, an operational profile of the bearing is determined based on the operational data. The operational profile includes a vibration response, a thermal response, and / or a frequency response associated with the bearing. For example, the operational data may include vibration sensor data. The vibration sensor data is used to generate vibration responses at various locations, such as radial locations outside the bearing, axial locations within the bearing, and radial locations within the bearing. The vibration responses may be generated along with a current signature to validate the generated vibration responses.

[0119] At step 706, an impact force profile is determined during bearing operation based on the operating profile and the virtual bearing model. Figure 3 The virtual bearing model is comparable to the model generated in

[15] . Thus, the virtual bearing model is generated based on a training model generated from the operating profile and impact force profile associated with a bearing set comparable to the same bearing. In one embodiment, the operating profile of the bearing set is superimposed on the operating profile of the (unknown) bearing. Based on this superposition, the impact force profile of the (unknown) bearing is determined from the impact force profile of the bearing set.

[0120] At step 708, a stress distribution associated with the bearing is predicted during bearing operation. The stress distribution is predicted based on the impact force profile. Additionally, the stress distribution can be generated directly from the virtual bearing model based on superposition.

[0121] At step 710, the remaining life of the bearing is predicted based on the stress distribution. A virtual bearing model is also used to determine the remaining life. For example, the virtual bearing model includes a predicted life based on the impact force profile of the bearing set. The remaining life is predicted based on the life predicted by the virtual bearing model.

[0122] At step 720, a degradation analysis is performed on the bearing by simulating various lubricant contamination conditions and varying lubricant viscosities on the virtual bearing model. For example, lubricant contamination conditions include normal contamination, mild contamination, severe contamination, and extreme contamination. For example, the degradation analysis can be used to determine bearing fatigue based on factors such as increased lubricant temperature, foreign particles in the lubricant, and decreased lubricant film parameters.

[0123] At step 730, the remaining life is updated using one or more neural networks configured to perform gradient descent optimization. The one or more neural networks include a simple neural network and a multivariate regression network.

[0124] In an embodiment, the simple neural network is applied using cross-entropy as a loss function. This is advantageous because bearing family information can not be readily determinable. Further, the operational data of the bearing can have non-linear relationships and can not be readily comparable. Bearing dimensions associated with the bearing are input to the multivariate regression network. The output of the one or more neural networks includes an updated remaining life based on the bearing dimensions and the bearing family parameters.

[0125] At step 740, a defect in the bearing is identified based on the impact force profile. Further, a contamination condition of the lubricant in the bearing is displayed on a display device.

[0126] Figure 8 FIG. illustrates an apparatus 820 for determining a condition of a bearing 810 in a system 800 at runtime, according to an embodiment of the present application.

[0127] The bearing 810 is connected to one or more sensing units 812. The sensing units 812 are used to measure operational parameters of the bearing 810 and the system 800. The measured operational parameters are referred to as operational data hereinafter. The operational data is input to the apparatus 820, which is configured to determine a condition of the bearing 810.

[0128] The apparatus 820 includes a processing unit 822, a communication unit 824, and a memory unit 825. In some embodiments, the apparatus 820 can include the sensing units 812. The apparatus 820 is communicatively coupled to a database 880 provided in a cloud computing environment via the communication unit 824 and a wireless communication network. The database 880 includes operational profiles 882 and impact force profiles 884 of a group of bearings comparable to the bearing 810.

[0129] The memory unit 825 includes machine readable instructions stored as modules, such as a virtual bearing module 830 and a bearing module 840. The modules 830 and 840 are executed by the processing unit 822 during runtime.

[0130] The virtual bearing module 830 includes bearing history data 832 and impact force profiles 834. The bearing history data 832 can include data associated with the design and manufacture of the bearing 810. The bearing history data 832 can further include catalog data and defect history. The impact force profiles 834 can include the impact force profiles 884 or a selection of the impact force profiles 884. The selection of the impact force profiles 884 is made based on the operational data from the system 800.

[0131] The bearing module 840 includes a pre-processing module 845, a response module 850, a machine learning module 860, and a condition module 870. Modules 845, 850, and 860 are described below.

[0132] During operation, operational data is received by the device and analyzed by pre-processing module 845. Pre-processing module 845 is configured to normalize the operational data to convert it into a suitable format for analysis. Response module 850 is configured to generate an operational profile of bearing 810 from the formatted operational data. In an embodiment, the operational profile includes a frequency-domain temperature response associated with the lubricant of bearing 810.

[0133] Machine learning module 860 includes a neural network and a regression network. The neural network uses cross-entropy as a loss function to determine the bearing family associated with bearing 810. Input to the neural network is catalog data associated with bearing 810 and bearing groups. Furthermore, the operating profile of bearing 810 is also input to the neural network.

[0134] The regression network is used to determine the size of the bearing 810 including rolling elements (i.e., balls). The regression network uses gradient descent optimization to perform multivariate regression on the operational data and thereby determine the size of the bearing 810. In addition, gradient descent optimization is used to determine the weights of the regression network.

[0135] The condition module 870 further analyzes the output of the machine learning module 860. The condition module 870 analyzes the output related to defects and stresses using the defect module 872 and the stress module 874, respectively.

[0136] In an embodiment, Figure 9 As shown in FIG, a virtual bearing model 830 of the device 820 is provided on a cloud computing platform.

[0137] Figure 9 Illustrated is a system 900 for determining the condition of a plurality of bearings 912, 922, and 932 in one or more systems 910, 920, and 930. Each of the systems 910, 920, and 930 is equipped with an apparatus 820. The apparatus 820 in this embodiment includes only a bearing module 840.

[0138] The virtual bearing module 830 is provided on a cloud computing platform 940, which is communicatively coupled to the device 820 via a network interface 950. As used herein, "cloud computing" refers to a processing environment that includes configurable computing physical and logical resources, such as networks, servers, storage, applications, services, etc., and data distributed over a network, such as the Internet. A cloud computing system provides on-demand network access to a shared pool of configurable computing physical and logical resources. The network is, for example, a wired network, a wireless network, a communication network, or a network formed by any combination of these networks.

[0139] The functions of the bearing module 840 and the virtual bearing module 830 to determine the condition of the bearings 912, 922, and 932 are as described above.

[0140] The system 900 can also include a display device 960 configured to display the remaining life of the bearings 912, 922, and 932. In embodiments, the defects in the bearings 912, 922, and 932 can also be displayed by superimposing the defects onto the systems 910, 920, and 930 using augmented reality technology.

[0141] While the application has been described in detail with reference to certain implementations thereof, it is understood that the application is not limited to those implementations. Rather, it will be understood that numerous modifications and variations can be resorted to without departing from the scope of the application as described herein. Accordingly, the scope of the application is indicated by the following claims, and their equivalents.

[0142] Figure 1

[0143] Stages 102-112

[0144] 102 "Normal Condition" of the bearing

[0145] 104 "Mild Contamination" of the lubricant

[0146] 106 "Severe Contamination" of the lubricant

[0147] 108 "Temperature Condition" of the lubricant

[0148] 110 "System Condition"

[0149] 112 "Load Condition"

[0150] Figure 2

[0151] Life Condition Relationship 210

[0152] Normal Contamination 212

[0153] Mild Contamination 214

[0154] Severe Pollution 216

[0155] Extreme Pollution 218

[0156] Lifetime 220

[0157] Remaining Lifetime 222, 224, 226, 228

[0158] Remaining Lifetime Under Normal Use 250

[0159] Table Column 260

[0160] Graph 270

[0161] Figure 3 Method

[0162] Figure 4

[0163] Virtual Bearing Model 400

[0164] Simulated Ball 402

[0165] Outer Raceway 404

[0166] Leading Edge 406

[0167] Trailing Edge 408

[0168] Leading Edge Force 412

[0169] Trailing Edge Force 414

[0170] Simulated Defect 450

[0171] Figure 5

[0172] Impact Cycle 500

[0173] Impulses 510 and 520

[0174] Figure 6

[0175] Virtual Bearing Model 600

[0176] Simulated Ball 602

[0177] Outer Raceway 604

[0178] Leading Edge 606

[0179] Trailing Edge 608

[0180] Graph 620

[0181] Leading Edge Force 622

[0182] Trailing Edge Force 624

[0183] Defect width 625

[0184] Simulation Defect 650

[0185] Figure 7 method

[0186] Figure 8

[0187] System 800

[0188] Device 820

[0189] Bearing 810

[0190] Sensing unit 812

[0191] Processing unit 822

[0192] Communication unit 824

[0193] Memory unit 825

[0194] Virtual Bearing Module 830

[0195] Historical data 832

[0196] Impact Profile 834

[0197] Bearing module 840

[0198] Preprocessing module 845

[0199] Response Module 850

[0200] Machine Learning Module 860

[0201] Status Module 870

[0202] Defective module 872

[0203] Stress Module 874

[0204] Database 880

[0205] Operation Profile 882

[0206] Impact Profile 884

[0207] Figure 9

[0208] System 900

[0209] Bearings 912, 922 and 932

[0210] Systems 910, 920, and 930

[0211] Cloud computing platform 940

[0212] Network interface 950

[0213] Display device 960.

Claims

1. A computer-implemented method of determining a condition of at least one bearing (810) in a system (800), the method comprising: receiving operational data associated with the system (800) from one or more sensing units associated with the system (800); determining an operational profile of the at least one bearing (810) from the operational data, wherein the operational profile comprises at least one of a vibration response and a thermal response associated with the at least one bearing (810); determining an impact force profile during operation of the at least one bearing (810) based on the operating profile and a virtual bearing model (400, 600), wherein the impact force profile is associated with an impact between rolling elements of the at least one bearing and a physical defect of the bearing, wherein the virtual bearing model (400, 600) is trained on operating profiles and impact force profiles associated with a group of bearings comparable to the at least one bearing (810) and is trained on determining simulated impact forces based on the test operating profile and the simulated operating profile; and A condition of the at least one bearing (810) is determined based on the impact force profile by generating a stress distribution based on the impact force profile.

2. The method of claim 1 , wherein the virtual bearing model (400, 600) is trained on operating profiles and impact force profiles associated with a group of bearings comparable to the at least one bearing (810) comprising: A test operation profile is determined based on test operation data associated with the bearing set, wherein the test operation includes operational data generated from the bearing set during testing of the bearing, and wherein the test operation profile includes a vibration response and a thermal response associated with the bearing set.

3. The method of claim 2 , wherein determining the test operation profile based on the test operation data associated with the bearing set comprises: creating one or more predetermined defects in one or more bearings in the bearing set; operating a system including the bearing under one or more system load conditions, wherein the system load conditions are indicative of a system load on the system; and A test operating profile associated with the bearing set is determined for a system load condition, wherein the test operating profile is generated based on test operating data received from sensing units located inside and outside each bearing in the bearing set.

4. The method according to any one of claims 1 to 3, further comprising: Generate the stress distribution associated with the bearing set based on the simulated impact forces.

5. The method of claim 4 , wherein determining the simulated impact force based on the test operation profile and the simulated operation profile comprises: A simulated impact force is predicted based on a simulated operating profile and at least one mass of a bearing ball, a damping coefficient and a stiffness associated with the bearing, wherein the simulated impact force includes at least one of a steady-state component from steady-state movement of the bearing and a dynamic component associated with a predetermined defect.

6. The method of claim 4, wherein generating a stress distribution associated with the bearing set based on the simulated impact force further comprises: comparing the test operating profile with the simulated operating profile; updating a simulated impact force based on the comparison; generating stress distributions based on the updated simulated impact forces; and The stress distribution is mapped to a predetermined defect using the at least one machine learning algorithm.

7. The method of claim 6, wherein updating the simulated impact force based on a comparison between the test operating profile and the simulated operating profile further comprises: An updated simulated impact force associated with the bearing set is calibrated based on a difference between the test operating profile and the simulated operating profile using at least one machine learning algorithm.

8. The method of claim 7, wherein the at least one machine learning algorithm is a differential evolution algorithm, and wherein using the at least one machine learning algorithm to calibrate an updated simulated impact force associated with a bearing set based on a difference between the test operating profile and the simulated operating profile comprises: Define the upper and lower stress limits for the bearing set; determining a possible stress distribution within an upper stress limit and a lower stress limit by one of a mutation and a recombination operation, wherein the possible stress distribution is determined for the difference; and The stress distribution is selected from the possible stress distributions using a continuous function optimization based on the impact force difference.

9. The method of claim 2, wherein determining a test operation profile based on test operation data associated with the bearing set further comprises: A life of the bearing when experiencing a predetermined defect is predicted based on at least one of a bearing load, a load zone, a bearing clearance, a lubrication viscosity, and lubricant contamination associated with one or more bearings in the bearing set.

10. The method according to any one of claims 1 to 3, further comprising: predicting a stress distribution associated with the at least one bearing (810) during operation of the at least one bearing (810), wherein the stress distribution is predicted based on an impact force profile of the at least one bearing (810) and a virtual bearing model (400, 600); and The remaining life of the at least one bearing (810) is predicted based on the stress distribution and the predicted life using a neural network; wherein the neural network is configured to perform gradient descent optimization.

11. The method according to any one of claims 1-3, wherein determining the condition of the at least one bearing (810) based on the impact force profile comprises: identifying defects in the at least one bearing (810) based on the impact force profile; and determining a contamination condition of the lubricant in the at least one bearing (810), Defects are identified and contamination conditions are determined by superimposing an operating profile of the at least one bearing (810) on an operating profile in a virtual bearing model (400, 600).

12. The method according to any one of claims 1-3, wherein determining the condition of the at least one bearing (810) based on the impact force profile further comprises: Fatigue of the at least one bearing (810) is determined based on the virtual bearing model (400, 600) with respect to at least one of a lubricant temperature increase, foreign particles in the lubricant, and a decrease in an oil film parameter of the lubricant.

13. The method according to claim 1 or 2, further comprising: simulating a predetermined defect on a predefined bearing model associated with the bearing set; as well as Based on the simulation of a predetermined defect on a predefined bearing model, a simulated operating profile associated with the bearing set is generated.

14. An apparatus for determining a condition of at least one bearing (810) in a system (800), the apparatus comprising: one or more (822) processing units; and a memory unit (825) communicatively coupled to the one or more processing units, wherein the memory unit includes a bearing module (840) and a virtual bearing module (830) stored in the form of machine-readable instructions executable by the one or more processing units, wherein the bearing module (840) is configured to perform one or more method steps according to claim 1 and any one of claims 10 to 12, and Wherein the virtual bearing module (830) is configured to perform one or more method steps according to any one of claims 2 to 9 or claim 13.

15. A system (800), comprising: one or more devices capable of providing operational data associated with the operation of one or more systems; and The apparatus of claim 14, communicatively coupled to the one or more devices, wherein the apparatus is configured to determine a condition of at least one bearing (810) in one or more systems according to any one of the method claims 1 to 13.

16. A computer program product having machine-readable instructions stored therein, which, when executed by one or more processing units, cause the one or more processing units to perform the method according to any one of claims 1 to 13.

Images