Method for determining the ageing state of a lubricant

Abstract

The present invention relates to a method for determining the ageing state of a lubricant, comprising: - a phase of constructing a training database (BD) using different lubricants; - a phase of training a model (M) according to the training database (BD) so as to obtain a trained model (M') in order to determine physico-chemical analysis data of a lubricant according to measurements of the characteristics of interest of the lubricant; - a phase of operating the trained model (M') for a target lubricant, comprising: · receiving measurements of characteristics of interest of the target lubricant, the measurements having been acquired by a sensor (C); · determining, by the trained model (M'), physico-chemical analysis data of the target lubricant; and - a phase of determining the ageing state of the target lubricant according to the physico-chemical analysis data determined.

Description

[0001] TITLE: Method for determining the aging state of a lubricant

[0002] The present invention relates to a method for determining the aging state of a lubricant. The present invention also relates to an associated computer program product.

[0003] Lubricants are used in industry for many applications, including agriculture, construction, and automotive.

[0004] To ensure the proper functioning of a lubricant, it is common to carry out physicochemical analyses on the lubricant to obtain analytical data. Such an analysis makes it possible to determine the aging state of the lubricant, in particular whether the lubricant retains its intrinsic properties essential for good lubrication, and whether the machine is in good working order, without excessive mechanical wear.

[0005] However, carrying out these analyses is time-consuming, and the reliability of the results can be affected by the multiple manipulations they involve (contamination during sample taking, possible errors during analysis, possible errors in writing the analysis report, etc.).

[0006] There is therefore a need for a tool that can determine the aging state of a lubricant in a simplified and reliable manner.

[0007] For this purpose, the present description relates to a method for determining the aging state of a lubricant, the method being implemented by computer and comprising:

[0008] - a phase of construction of a training database from different lubricants, the construction phase comprising:

[0009] • obtaining initial data for each lubricant, the initial data being measurements of a set of characteristics, called characteristics of interest, of the lubricant, the measurements having been acquired by at least one sensor,

[0010] • processing of the first data so as to compensate for temperature variations of at least one sensor,

[0011] • obtaining second data for each lubricant, the second data being derived from a physicochemical analysis carried out on the lubricant,

[0012] • the construction of the training database (DB) by association, for each lubricant, of the first processed data with the second data,

[0013] - a phase of training a model based on the training database so as to obtain a trained model to determine physicochemical analysis data of a lubricant based on measurements of the characteristics of interest of the lubricant, and

[0014] - an exploitation phase of the model trained for a target lubricant, comprising the following steps:

[0015] • receiving measurements of characteristics of interest of the target lubricant, the measurements having been acquired by at least one sensor,

[0016] • determination by the trained model of physicochemical analysis data of the target lubricant,

[0017] - a phase of determining the aging state of the target lubricant based on the determined physicochemical analysis data.

[0018] According to particular embodiments, the method comprises one or more of the following characteristics, taken in isolation or in all technically possible combinations:

[0019] - processing of the first data also includes the elimination of outliers in the first data;

[0020] - the processing of the first data also includes the filtering of the first data on the domain of validity of the application for which the lubricant is used and / or the domain of validity of the at least one sensor;

[0021] - the phase of determining the aging state of the target lubricant comprises the determination of a desirability score for each determined physicochemical analysis data, the aging state of the target lubricant being determined according to the desirability scores;

[0022] - the phase of determining the aging state of the target lubricant also includes the determination of an overall desirability score based on all of the determined desirability scores, the overall desirability score being representative of the aging state of the target lubricant;

[0023] - the phase of determining the aging state of the target lubricant also includes the generation of an alert or a recommendation by comparing the desirability scores or the overall desirability score to a predetermined threshold;

[0024] - the trained model is specific to a category of lubricants, the method comprising repeating the construction and training phases for other categories of lubricants, so as to obtain a library of trained models for different categories of lubricants, the method comprising, before the exploitation and aging state determination phases, a phase of selecting a trained model from the library of models as a function of the target lubricant considered; - at least one characteristic of interest is chosen from the following characteristics of the lubricant: temperature, viscosity, density and dielectric constant;

[0025] - the sensors are installed on operating equipment, such as an engine, a hydraulic circuit, or a gearbox, so as to enable monitoring of the aging state of the target lubricant in real time.

[0026] The present description also relates to a computer program product comprising program instructions recorded on a computer-readable medium, for executing a method as described above when the computer program is executed on a computer.

[0027] The present description also relates to a readable information medium on which a computer program product as previously described is stored.

[0028] Other characteristics and advantages of the invention will appear on reading the following description of embodiments of the invention, given by way of example only and with reference to the drawings which are:

[0029] [Fig 1], Figure 1, a schematic view of an example of a computer enabling the implementation of a method for determining the aging state of a lubricant, [Fig 2], Figure 2, a flowchart of an example of implementation of a method for determining the aging state of a lubricant, and

[0030] [Fig 3], Figure 3, a schematic representation of an example of the constitution of a database used for training a model capable of determining physicochemical analysis data of a lubricant based on measurements of the characteristics of interest of the lubricant.

[0031] A calculator 10 and a computer program product 12 are illustrated in Figure 1.

[0032] The calculator 10 is preferably a computer.

[0033] More generally, the computer 10 is an electronic computer capable of manipulating and / or transforming data represented as electronic or physical quantities in computer 10 registers and / or memories into other similar data corresponding to physical data in memories, registers or other types of display, transmission or storage devices.

[0034] The calculator 10 interacts with the computer program product 12.

[0035] As illustrated in FIG. 1, the computer 10 comprises a processor 14 comprising a data processing unit 16, memories 18 and an information medium reader 20. In the example illustrated in FIG. 1, the computer 10 comprises a keyboard 22 and a display unit 24.

[0036] The computer program product 12 comprises an information medium 26. The information medium 26 is a medium readable by the computer 10, usually by the data processing unit 16. The readable information medium 26 is a medium suitable for storing electronic instructions and capable of being coupled to a bus of a computer system.

[0037] For example, the information medium 26 is a floppy disk or flexible disk (also known as a “Floppy disk”), an optical disk, a CD-ROM, a magneto-optical disk, a ROM memory, a RAM memory, an EPROM memory, an EEPROM memory, a magnetic card or an optical card.

[0038] The computer program 12 comprising program instructions is stored on the information medium 26.

[0039] The computer program 12 is loadable onto the data processing unit 16 and is adapted to cause the implementation of a method for determining the aging state of a lubricant, when the computer program 12 is implemented on the processing unit 16 of the computer 10.

[0040] The operation of the calculator 10 will now be described with reference to Figure 2, which schematically illustrates an example of implementation of a method for determining the aging state of a lubricant, and to Figure 3 which is an example of a phase of the method.

[0041] The determination method comprises a phase 100 of constructing a training database BD from different lubricants. Phase 100 is implemented by the computer 10 in interaction with the computer program product 12, i.e. is implemented by computer.

[0042] Phase 100 comprises a step 110 of obtaining first data D1 for each lubricant. The first data D1 are measurements of a set of characteristics, called characteristics of interest, of the lubricant. The measurements were acquired by at least one sensor C.

[0043] Preferably, at least one characteristic of interest is chosen from the following characteristics of the lubricant: temperature T, viscosity V, density D and dielectric constant DC.

[0044] For example, as illustrated in Figure 3, the characteristics of interest were measured by one or more sensors C installed on operating equipment E, such as an engine, a hydraulic circuit, or a gearbox.

[0045] C-sensors are, for example, connected sensors.

[0046] For example, the first data D1 are measured at regular time intervals by the sensor(s) C, for example every 30 seconds. Phase 100 comprises a step 120 of processing the first data D1 to obtain first processed data D1'.

[0047] The processing makes it possible in particular to compensate for the temperature variations of at least one sensor C. A smoothing filter is preferably applied to the temperature-compensated data.

[0048] More particularly, processing step 120 consists of determining values ​​for the characteristics of interest corrected for the effect of temperature (and measurement biases), for example, the corrected viscosity V*, the corrected density D* and the corrected dielectric constant DC*.

[0049] In an example of implementation, the temperature curves of each sensor C are measured, or provided in manufacturer data. The temperature sensitivity (slopes) of the characteristics of interest considered is thus estimated. A compensation is then applied to the first data D1 to compensate for the temperature effect (for example by placing the sensor at a reference temperature).

[0050] Alternatively, a model is trained (machine learning or neural network) to determine the characteristics of interest (V*, D*, DC*) corrected for the effect of temperature based on the characteristics of interest (V, D, DC) and the temperature considered.

[0051] Preferably, the processing of the first data D1 also comprises the elimination of outliers in the first data D1. The eliminated data are, for example, data outside the validity range of the sensor C. The validity range is, for example, given by the manufacturer and consists of acceptable value intervals, beyond which the manufacturer does not guarantee the reliability of the data.

[0052] Preferably, the processing of the first data D1 also includes filtering the first data D1 on the domain of validity of the application for which the lubricant is used. For example, the measurements taken during engine shutdown are eliminated because the circulation of the lubricant is no longer ensured at the stop.

[0053] Preferably, the processing of the first data D1 also comprises the normalization of the first data D1 with respect to the invariants of the respective signal from which the first data D1 originate. An invariant is for example the minimum value over an interval or certain quantiles of the distribution of the signal values.

[0054] Phase 100 comprises a step 130 of obtaining second data D2 for each lubricant. The second data D2 come from a physicochemical analysis carried out on the lubricant. The second data D2 have typically been carried out in a laboratory. Preferably, the second data D2 have also been preprocessed, for example so as to remove outliers or correct biases in the data.

[0055] The second D2 data are, for example, obtained at regular intervals, for example once a week.

[0056] The second data D2 are, for example, chosen from the following data:

[0057] - Standard analyses (Noack, flash point, pour point, cloud point, foaming, density, ash content, flash point, etc.).

[0058] - Kynematic viscosities (KV40, KV100, VI).

[0059] - Gas chromatography (fuel content).

[0060] - IR spectroscopy (soot content, oxidation, nitration).

[0061] - Titration (TAN, TBN, ipH, water content).

[0062] - Element content (ICP).

[0063] Phase 100 includes a step 140 of constructing the training database BD by associating, for each lubricant, the first processed data DT with the second data D2 (see figure 3).

[0064] For example, in the case where the first processed data DT are more numerous than the second data D2, the association is carried out on the basis of the dates (i.e. the first sensor data for which second data D2 are available).

[0065] The determination method comprises a phase 200 of training a model M according to the training database BD so as to obtain a trained model M' (see FIG. 3) for determining physicochemical analysis data of a lubricant according to measurements of the characteristics of interest of the lubricant. Phase 200 is implemented by the computer 10 in interaction with the computer program product 12, that is to say is implemented by computer.

[0066] The model is, for example, a neural network. The learning technique allows the neural network to be configured as it is trained on the database (supervised learning). In one example, part of the database is used to configure the neural network and the other part to validate the configuration.

[0067] Alternatively, the model implements one of the following algorithms: Random Forest, XGboost Regressor, Gradient Boosting, Linear Regression, Ridge Regression (alpha=1 ), and SVM (with linear kernel and radial basis function).

[0068] The determination method comprises a phase 300 of operating the trained model M' for a target lubricant. The phase 300 is implemented by the computer 10 in interaction with the computer program product 12, that is to say is implemented by computer.

[0069] Phase 300 comprises receiving measurements of characteristics of interest of the target lubricant. The measurements were acquired by at least one sensor C. The data acquired by the at least one sensor C are advantageously processed to compensate for temperature variations of the at least one sensor C, in a manner similar to what was described for the data D1 used during phase 100.

[0070] Preferably, the C-sensors are installed on operating equipment E, such as an engine, a hydraulic circuit, or a gearbox, so as to enable monitoring of the aging state of the target lubricant in real time. The C-sensors are, for example, connected sensors.

[0071] Phase 300 then comprises the determination by the trained model M' of the state of physicochemical analysis data of the target lubricant.

[0072] The determination method comprises a phase 400 of determining the aging state of the target lubricant based on the determined physicochemical analysis data. Phase 400 is implemented by the computer 10 in interaction with the computer program product 12, i.e. is implemented by computer.

[0073] In one example, phase 400 comprises determining a desirability score for each determined physicochemical analysis data. The desirability score is, for example, between 0 and 1.

[0074] For example, the desirability score is obtained using the following equation:

[0075] Or :

[0076] Ÿ t denotes the characteristic of interest i to be predicted,

[0077] Lt denotes the minimum value to be set for the characteristic of interest i, and

[0078] Ti denotes the maximum value to be set for the characteristic of interest i.

[0079] In a variant, d t is not bounded to 1 and values ​​greater than 1 are allowed for d t .

[0080] Preferably, step 400 also comprises determining an overall desirability score based on all of the determined desirability scores. The overall desirability score is representative of the aging state of the target lubricant.

[0081] For example, the total desirability score is obtained using the following equation:

[0082] Where k denotes the number of characteristics of interest considered. The desirability score is typically expressed as a percentage.

[0083] Optionally, a different weight is applied to the scores of the different characteristics of interest when calculating the total desirability score. The different weights, minimum values, maximum values, are determined for each application.

[0084] Preferably, phase 400 also comprises generating an alert or recommendation by comparing the desirability scores or the overall desirability score to a predetermined threshold. For example, an alert or recommendation is made when the desirability score is greater than 85% (0% designating a least used lubricant condition and 100% a most used lubricant condition). This indicates, for example, that maintenance is required.

[0085] Preferably, the trained model is specific to a category of lubricants, the method comprising repeating the construction 100 and training 200 phases for other categories of lubricants, so as to obtain a library of trained models for different categories of lubricants. The method comprises, in this case, before the exploitation 300 and determination 400 phases, a phase 125 of selecting a trained model from the library of models as a function of the target lubricant considered. The selection is, for example, carried out by a user, or automatically as a function of the type of the lubricant category.

[0086] Thus, once the model has been trained, the present method makes it possible to determine in a simplified manner the aging state of a lubricant, without physicochemical analyses and with only measurements from C sensors. The model having been trained on the basis of data taking into account the temperature variations of the sensor, the reliability is improved.

[0087] This method thus makes it possible to prevent failures of a lubricant in operation, to help plan equipment maintenance, and to replace the lubricant only when it is faulty or aging.

[0088] Those skilled in the art will understand that the previously described embodiments and variations may be combined to form new embodiments provided that they are technically compatible.

Claims

CLAIMS 1. Method for determining the aging state of a lubricant, the method being implemented by computer and comprising: - a phase of construction of a training database (DB) from different lubricants, the construction phase comprising: • obtaining first data (D1) for each lubricant, the first data (D1) being measurements of a set of characteristics, called characteristics of interest, of the lubricant, the measurements having been acquired by at least one sensor (C), • processing of the first data (D1) so as to compensate for temperature variations of at least one sensor (C), • obtaining second data (D2) for each lubricant, the second data (D2) being derived from a physicochemical analysis carried out on the lubricant, • the construction of the training database (DB) by association, for each lubricant, of the first processed data (DT) with the second data (D2), - a phase of training a model (M) based on the training database (BD) so as to obtain a trained model (M') to determine physicochemical analysis data of a lubricant based on measurements of the characteristics of interest of the lubricant, - an exploitation phase of the trained model (M') for a target lubricant, comprising the following steps: • receiving measurements of characteristics of interest of the target lubricant, the measurements having been acquired by at least one sensor (C), • the determination by the trained model (M') of physicochemical analysis data of the target lubricant, - a phase of determining the aging state of the target lubricant based on the determined physicochemical analysis data.

2. The method of claim 1, wherein processing the first data (D1) also includes removing outliers from the first data (D1).

3. Method according to claim 1 or 2, in which the processing of the first data (D1) also comprises the filtering of the first data (D1) on the domain of validity of the application for which the lubricant is used and / or the domain of validity of the at least one sensor (C).

4. Method according to any one of claims 1 to 3, in which the phase of determining the aging state of the target lubricant comprises the determination of a desirability score for each determined physicochemical analysis data, the aging state of the target lubricant being determined as a function of the desirability scores.

5. Method according to claim 4, in which the phase of determining the aging state of the target lubricant also comprises determining an overall desirability score based on all of the determined desirability scores, the overall desirability score being representative of the aging state of the target lubricant.

6. Method according to claim 5, in which the phase of determining the aging state of the target lubricant also comprises the generation of an alert or a recommendation by comparing the desirability scores or the overall desirability score to a predetermined threshold.

7. Method according to any one of claims 1 to 6, in which the trained model (M') is specific to a category of lubricants, the method comprising the repetition of the construction and training phases for other categories of lubricants, so as to obtain a library of trained models for different categories of lubricants, the method comprising, before the exploitation and aging state determination phases, a phase of selecting a trained model (M') from the library of models as a function of the target lubricant considered.

8. Method according to any one of claims 1 to 7, in which at least one characteristic of interest is chosen from the following characteristics of the lubricant: temperature, viscosity, density and dielectric constant.

9. Method according to any one of claims 1 to 8, in which the sensors (C) are installed on equipment (E) in operation, such as an engine, a hydraulic circuit, or a gearbox, so as to enable the monitoring of the aging state of the target lubricant in real time.

10. A computer program product comprising program instructions recorded on a computer-readable medium, for executing a determination method according to any one of claims 1 to 9 when the computer program is executed on a computer.

Images