METHOD FOR ESTIMATING THE WEAR CONDITION OF A TIRE
Patent Information
- Application Number
- DE602021021468
- Authority / Receiving Office
- DE · DE
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2020-08-03
- Filing Date
- 2021-07-28
- Publication Date
- 2024-11-06
- Estimated Expiration
- 2041-07-28
AI Technical Summary
Existing methods for estimating tire wear using vehicle-mounted systems are complex, resource-intensive, and difficult to implement in real-time due to the high number of parameters involved, and are affected by indirect considerations such as driving speed, making them costly and inefficient, especially for entry-level vehicles.
A method that records vibroacoustic signals from tires, transforms them into frequency signals, segments these signals into bands, and uses automatic learning based on a learning database to predict tire wear, incorporating driving speed and ground conditions to reduce calculation times and costs while improving robustness.
This approach allows for quick, accurate, and cost-effective estimation of tire wear in real-time, reducing the need for extensive resources and simplifying implementation, while enhancing the method's robustness by accounting for driving speed and ground conditions.
Description
Field of invention
[0001] The present invention relates to a method for estimating the state of wear of a tire fitted to a vehicle moving on a road, from the noise generated by the tire when it is in dynamic contact with the ground. Technological background
[0002] It is useful to know the state of wear of the tire to interact with the driver or the driving assistance systems, so as to inform them in real time of the modification of the grip conditions of the tire and the road holding of the vehicle by the modification in particular of the state of wear of the latter.
[0003] Thus, patent applications FR3015036A1 and WO2017103474A1 propose a method that combines both the state of wear of the tire and the condition of the ground on which the tire rolls. However, this method, which is possible in on-board vehicle conditions, implements a large number of parameters to be managed by discriminant analysis. This high number of parameters leads to calculation times and calculation resources that make this method complicated to implement for an acceptable processing cost regardless of the type of motor vehicle, and in particular, entry-level vehicles. In addition, the robustness of the method is also affected by indirectly taking into account the influence of the driving speed during data processing.
[0004] One of the objects of the invention which follows aims to propose a method making it possible to highlight the evolution of the state of wear of the tire by an autonomous method, that is to say without human intervention and without immobilization of the vehicle. In addition, the method must be easy to implement, with reduced costs and calculation times, to facilitate real-time processing, particularly when the post-processing of the data is carried out at the vehicle level. Description of the invention
[0005] The invention relates to a method for estimating the state of wear of a tire of a mounted assembly of a vehicle rolling on a road surface comprising the following steps: Recording a measurement of a vibroacoustic signal produced by the tire rolling on the road surface during a given time frame; Transforming the time signal into a frequency signal over a given frequency interval; Segmenting the frequency interval into at least two frequency bands of predetermined width and associating with each at least two frequency bands at least one data item representative of the frequency signal in said at least one frequency band, the at least one representative data item resulting from the measurement forming at least one variable of a matrix associated with said measurement;Predict a tire wear condition corresponding to the matrix associated with the measurement carried out, using automatic data learning based on a learning base formed from a set of matrices associated with recorded measurements and carried out according to the same steps as above, in known driving conditions according to modalities each representing a tire wear condition; and Determine the tire wear condition after a number N of identical forecasts in a series M of consecutive forecasts. ;
[0006] The term "vibroacoustic signal" here means any vibratory and / or acoustic signal that can be captured by a sensor, for example a microphone or accelerometer.
[0007] The idea is to use a sensor that does not significantly modify the response of the tire rolling on the road surface when it is mounted on a vehicle via a rim for example. The rolling condition of the tire on the road generates a specific signature, whether vibratory and / or acoustic, in a wide frequency spectrum. The principle consists of measuring a time signal over a short time and transforming it into a frequency signal for which the useful information is more accessible since it is distributed over a wide frequency band. Assuming that the signal is pseudo-periodic, it is easily possible to transform a small time sample with a certain sampling into a wideband frequency spectrum with a given frequency step. The sampling frequency determines the width of the frequency spectrum.Once this spectrum is obtained, it should be separated into a multitude of frequency bands, each band representing a dimension of the matrix associated with the vibroacoustic signature. Then, for each band, one or more data representative of the frequency spectrum on the frequency band are identified. These representative data represent the second dimension of the matrix. These data representative of the frequency spectrum are included in the group comprising the mean, the median, the maximum value, the minimum value or a combination of these quantities of the frequency spectrum. These representative data can also be quantities relating to the shape of the spectrum on the frequency band which are translated by the emergences on a frequency or a band of frequencies compared to the average level on the frequency band. The division into frequency bands for the dimension of the matrix can for example be carried out by third of an octave.
[0008] Finally, the prognosis of the tire wear condition is carried out by automatic learning based on matrices built on the model of those obtained using the matrix formed in the previous step during a learning campaign. During this learning campaign, various tire runs were carried out during which a vibroacoustic measurement was recorded and then analyzed according to the same protocol for each run. Each run was carried out on a specific vehicle with a known tire assembly including in particular the level of tire wear under various conditions of pressure and static load applied around the nominal conditions. Each run consisted of driving the vehicle on a road whose macrorough nature was predetermined, the meteorological state of the road was identified.For this purpose, the state of wear of the tire was divided into at least two groups called new or worn according to the height of the tire tread. When the height of the tread for a tire mounted inflated on a rim, between the radially outer height of the tread in the new state, called maximum height, and the minimum height corresponding to the radially outer distance of the wear indicator at the bottom of the grooves of the tread, is located between the maximum height and the maximum mid-height which is equidistant from the maximum height and the minimum height, the state of the tire is said to be "new" and conversely, it is said to be "worn". The state of wear can also be separated into a multitude of classes of identical size.For example, if we want to classify the state of wear of the tire as "new", "half-worn" and "worn", the dividing lines between these various states are reciprocally two-thirds of the difference between the maximum height and the minimum height and one-third of the difference in heights. Thus each class or group is distributed over a band of similar tread heights.
[0009] Here, the multitude of measurements and forecasts of the tire wear condition allows statistically to quickly evaluate the tire wear class through a simple model. This simple mathematical model requires only a small learning base and generates a very fast response time. Thus, it is assumed that the N occurrences of the same forecast in a series of M measurements, for example whatever, for example, the meteorological state of the road, the macrorough nature of the road or the driving speed will make it possible to detect the transition from one tire wear class to another. In fact, the tire wear condition evolves slowly in the time domain compared to the other sensitive quantities of the system. The same applies to the tire usage conditions through parameters such as pressure or the applied static load.Of course, the series of measurements M will be dictated by the diversity of the measurement conditions and the number of occurrences N will be driven by the accuracy of the prognosis in said measurement conditions. Thus, unlike the model in the state-of-the-art document, we choose to multiply the measurements on an elementary model to predict, by information redundancy, the state of the tire which by nature evolves slowly. The simplicity of the model allows for fast response times and post-processing on board the vehicle. The prediction also gains in robustness by not putting under control too large a set of sensitive parameters of the vibroacoustic signature of the tire.
[0010] Preferably, the method comprises the following step: Determine a tire rolling speed class whose width is a fraction of a maximum rolling speed, preferably the maximum rolling speed is 300 km / h; and in which the rolling speed class is a modality of machine learning.
[0011] In this particular case, the rolling speed class is added as a modality to the learning base. This allows for faster prediction of the tire wear condition since the rolling speed largely influences the vibroacoustic response of the rolling tire, particularly at the level of the average level measured over a wide frequency window. This influence is reflected in particular by an increase in the vibroacoustic response of the tire proportional to the rolling speed. The sensitivity of the representative data may be reduced or disappear in the vibroacoustic response driven by the average level. Thus, the sensitivity of the data representative of the frequency signal is enhanced by taking into account the rolling speed as a modality, which allows for improved prediction of the tire wear condition.Speed can be evaluated by data from the vehicle such as information from a GPS (acronym for Global Positioning System) or information transmitted on the vehicle's CANbus or information contained directly or indirectly through electronic systems mounted on the tire such as a TMS (acronym for Tire Mounted Sensor) or TPMS (acronym for Tire Pressure Monitoring System) or by any other means. For the simplicity of the system and due to the sensitivity of the vibroscoustic response of the tire, a classification of the speed by speed class is sufficient, preferably a speed class width of around ten km / h is a good compromise between the desired evaluation speed reduced to the size of the learning base and the desired accuracy on the tire wear condition. This reduces the learning base for prediction without too much impact on accuracy.In addition, the evolution of the tire wear condition is slow. Therefore, the redundancy of the wear condition information leads to statistically determining the tire wear condition from a certain number M of vibroacoustic measurements. The addition of the rolling speed or the rolling speed class makes the process more efficient in terms of the number of vibroacoustic measurements to converge towards the correct information on the tire wear condition.
[0012] According to a preferred embodiment, the determination of the vehicle's rolling speed comprises the following steps: Recording a second measurement of a vibroacoustic signal produced by the tire rolling on the road surface during a second given time frame; Transforming the second time signal into a second frequency signal over a second given frequency interval; Segmenting the second frequency interval into at least two frequency bands of predetermined width and associating with each at least two frequency bands at least one data item representative of the second frequency signal in said at least one frequency band, the at least one representative data item from the second measurement forming at least one variable of a matrix associated with said second measurement;Determine a tire rolling speed class corresponding to the matrix associated with the second measurement carried out, using a second automatic learning of the data based on a learning base formed from a set of matrices associated with recorded measurements and carried out according to the same steps as above, in known rolling conditions according to modalities each representing a tire rolling speed class.;
[0013] The evaluation of the rolling speed can be carried out through a vibroacoustic measurement of the tire rolling on the surface of a road. For this, the use of a second automatic learning built on a learning base taking into account as modality only the rolling speed or classes of rolling speeds is recommended. Indeed, the vibroacoustic response of the tire depends on the state of wear of the tire but also on other parameters. However, the rolling speed influences the entire frequency spectrum of the tire and not only certain specific frequency bands or at least these bands are not all identical according to the observed parameters. Thus, the data representative of the vibroacoustic signature matrix linked to the rolling speed is different from that of the state of wear of the tire.Therefore, it is possible by specific machine learning to identify the rolling speed or the rolling speed class corresponding to a vibroacoustic measurement of the tire by dividing the frequency signal into specific frequency bands and evaluating particular representative data on each frequency band. This frequency division and its representative data are by nature different from those associated with the matrix for identifying the state of wear of the tire. Of course, this second vibroacoustic measurement may be different from the vibroacoustic measurement of the state of wear. However, nothing prevents the same measurement from being used for both machine learning processes. In this particular embodiment, it is necessary to identify the rolling speed or the rolling speed class before determining the state of wear of the tire, which imposes the order of the post-processing.Finally, the division of the rolling speed into various speed classes makes it possible to limit the number of modalities, which makes the method more efficient both in terms of the size of the learning base and the response time of the post-processing. Unlike the state-of-the-art document, it is possible at the level of the evaluation of the rolling speed to take into account a series of representative data, unlike the only data corresponding to the average power on a given frequency band, which makes this method more robust in terms of precision and robustness. Thus, the number of measurements necessary to statically predict the state of wear of the tire is significantly reduced.
[0014] According to the invention, the method comprises the complementary steps: Determine a ground surface condition class; and wherein the ground surface condition class is a condition for carrying out the prognosis of the tire wear condition if it is a specific ground condition class.
[0015] Here, the ground surface condition class becomes a condition for carrying out the tire wear condition prognosis step. In the first case, the ground surface condition classes are added as a modality. This has the advantage of statistically accelerating the prediction of the tire wear condition by reducing the number of measurements M to be carried out to converge towards the useful information. However, the learning base becomes more substantial since these new modalities increase the number of combinations between the modalities. Thus, the size of the learning base is increased and the mathematical model associated with the tire wear condition is made more complex. In the second case, the inventor found that when the ground surface condition class is a specific ground condition class, the determination of the tire wear condition does not need to know the ground surface condition class to converge quickly towards the solution.Thus, the ground condition class becomes just an indicator to launch the prognosis of the tire wear condition. This second case makes it possible to limit the size of the learning base and the mathematical model of the tire wear condition. However, the number of measurements eligible for the prognosis is reduced to only those measurements where the condition on the specific ground condition class is obtained. This is not penalizing in the case of the tire wear condition which is a parameter that evolves slowly over time. Of course, it is necessary that the determination of the ground surface condition class be carried out before the prognosis step of the tire wear condition. It is entirely possible, but not essential, to combine the use of the ground surface condition class with the use of the rolling speed class to further reduce the number of measurements necessary to converge on the information on the tire wear condition.Finally, this state of the road surface can be obtained via meteorological information linked to a map associated with the GPS position of the vehicle or by any other means on the vehicle such as a rain sensor, a windshield wiper actuator, etc.
[0016] Specifically the ground surface condition class is included in the group including dry, humid, wet, snowy and icy.
[0017] The ground condition corresponds to the meteorological condition of the ground. The meteorological condition is included in the group comprising: a dry condition, a humid condition and a wet condition or a winter condition such as a snow or ice condition. A snow condition may, in a preferred embodiment, be included in the group comprising a fresh snow condition, a compacted snow condition, a granular snow condition and a slush condition.
[0018] A wet state is characterized by a water level that is flush with the natural roughness of the road surface. This wet state corresponds, for example, to the state of the ground obtained by a light rain or a road that dries out after a heavy rain. A wet state, on the other hand, is characterized by a water level that exceeds the level of the natural roughness of the road surface. In practice, the wet state corresponds to a water level generally between 0.5 millimeters and 1 millimeter.
[0019] The weather condition of the ground can be determined by weather forecasts associated with the GPS position of the vehicle, the response of sensors on the vehicle such as the actuation of the windshield wipers or a rain detection sensor on the windshield. Finally, by associating the ambient temperature of the environment in which the vehicle is moving, it is possible to identify the presence or formation of snow associated with a winter condition.
[0020] According to the invention, the specific soil condition class comprises the so-called "dry" soil surface condition class.
[0021] Among the specific meteorological conditions, which makes it possible not to take the ground condition class as a modality of the automatic learning linked to the state of wear of the tire, we will note the so-called dry ground surface condition for which the contact between the tire and the ground is not impacted by the presence of water whatever its form and whatever its quantity. The inventor found that the specific ground condition class could also mix the soils of the so-called "wet" and "dry" ground surface condition classes without harming the prognosis on the state of wear of the tire by the automatic learning although the restriction to only the so-called "dry" ground surface condition class improves the probability of the prognosis. However, it is preferable to avoid the so-called "wet" ground surface condition class.
[0022] This specific condition allows the selection of vibroacoustic measurements made in dry weather, regardless of the ambient temperature, which limits the number of measurements used to predict the tire wear condition, making the method more efficient. Indeed, the presence of water modifies part of the frequency spectrum obtained compared to that obtained using a vibroacoustic measurement in dry weather, which naturally creates dispersion in the frequency spectra and inevitably impacts the quality of the prognosis.
[0023] According to a second embodiment, the determination of the ground surface condition class comprises the additional steps: Recording a third measurement of a vibroacoustic signal produced by the tire rolling on the road surface during a third given time frame; Transforming the third time signal into a third frequency signal over a third given frequency interval; Segmenting the third frequency interval into at least two frequency bands of predetermined width and associating with each at least two frequency bands at least one data item representative of the third frequency signal in said at least one frequency band, the at least one representative data item from the third measurement forming at least one variable of a matrix associated with said third measurement;Determine a ground surface condition class corresponding to the matrix associated with the third measurement carried out, using a third automatic data learning based on a learning base formed from a set of matrices associated with recorded measurements and carried out according to the same steps as above, in known driving conditions according to modalities each representing at least one ground surface condition class.;
[0024] The evaluation of the meteorological state of the road surface can be carried out through a vibroacoustic measurement of the tire rolling on the surface of a road. For this, the use of a third machine learning built on a learning base taking into account as a modality the class of state of the ground. In addition, the class of rolling speed can also be taken into account as a modality of this third machine learning or as a condition for carrying out the determination of the state of the ground surface in order to reduce the learning base of this third machine learning, which makes the identification of the class of meteorological state of the road surface more efficient. Indeed, the vibroacoustic response of the tire depends on the state of wear of the tire but also on other parameters such as the rolling speed.However, the rolling speed influences the entire frequency spectrum of the tire and not only certain specific frequency bands or at least these bands are not all identical according to the observed parameters. Thus, the representative data of the vibroacoustic signature matrix linked to the meteorological state of the road is different from that of the tire wear state. Therefore, it is possible by specific machine learning to identify the class associated with the meteorological state of the road corresponding to a vibroacoustic measurement of the tire by cutting the frequency signal into specific frequency bands and evaluating particular representative data on each frequency band. This frequency cutting and its representative data are by nature different from those associated with the identification matrix of the tire wear state and the rolling speed.However, having previously identified the rolling speed class, it is possible to take this parameter into account to reduce the learning base of this third automatic learning, which makes the method more efficient. Of course, this third vibroacoustic measurement can be different from the vibroacoustic measurement of the state of wear and the vibroacoustic measurement of the rolling speed class. However, nothing prevents using the same vibroacoustic measurement as one or other of the measurements associated respectively with the other two automatic learnings, which reduces the number of steps to be carried out.
[0025] In this particular embodiment, it is necessary to identify the ground surface condition class before determining the tire wear condition, which imposes a certain order of post-processing. On the other hand, it is mandatory to carry out the post-processing of the vibroacoustic measurement associated with the ground condition class after the post-processing of the vibroacoustic measurement of the rolling speed class to take this parameter into account in the third automatic learning.
[0026] According to the invention, the method comprises the complementary steps: Determine a soil type class; and in which the soil type class is a condition for carrying out the prognosis of the tire wear condition if it is a specific soil type class.
[0027] Here, the soil type class becomes a condition for carrying out the tire wear prediction step. In the first case, the soil type class is added as a modality. This has the advantage of statistically accelerating the tire wear prediction by reducing the number of measurements to be made to converge towards the useful information. However, the learning base becomes more substantial since these new modalities increase the number of combinations between the modalities. Thus, the size of the learning base is increased and the mathematical model associated with the tire wear is made more complex. In the second case, the inventor found that when the soil type class is a specific soil type class, determining the tire wear does not need to know the soil type class to quickly converge towards the solution.Thus, the soil type class becomes just an indicator to launch the prognosis of the tire wear condition. This second case makes it possible to limit the size of the learning base and the mathematical model of the tire wear condition. However, the number of measurements eligible for the prognosis is reduced to only those measurements where the condition on the specific soil type class is obtained. This is not penalizing in the case of the tire wear condition which is a parameter that evolves slowly over time. Of course, it is necessary that the determination of the soil type class be carried out before the step of determining the tire wear condition. It is entirely possible, but not essential, to combine the use of the soil type class with the rolling speed class or the soil surface condition class to further reduce the number of useful measurements to converge on the information on the tire wear condition.Finally, this soil nature class can be obtained through soil mapping associated with the GPS position of the vehicle or by any other means such as optical and / or light and / or sound means on board the vehicle.
[0028] Specifically, the soil nature class is included in the group including “open”, “medium” and “closed”.
[0029] A surface is described as "closed" or "macro-smooth" when it takes on a smooth appearance without roughness, such as bitumen that has oozed after being subjected to high heat or a concrete slab covered with a layer of cement. A surface will be considered "open" or "macro-rough" when the roughness is significant, for example, such as that of a worn surface or that of a country road quickly repaired using a surface coating made by spraying pebbles onto bitumen. A "medium" surface describes all surfaces in an intermediate state between the two previous states and more specifically describes new surfaces.
[0030] The macro-texture of a coating affects the drainability properties of the road but also the acoustic properties of the tire. This texture is characterized by a "Mean Texture Depth", also called "PMT". This PMT is measured by the height to true sand, also noted HSv, well known to those skilled in the art.
[0031] The different macro-textures can be categorized as follows: a closed macro-texture coating has a PMT of between 0 and 0.4 millimeters. A medium macro-texture coating has a PMT of between 0.4 and 1.1 millimeters, and an open macro-texture coating has a PMT greater than 1.1 millimeters.
[0032] According to the invention, the specific soil nature class includes the so-called “open” soil nature class.
[0033] Among the specific soil nature classes, which makes it possible not to take the soil nature as a modality of automatic learning linked to the state of wear of the tire, we will note the soil nature class characterized by a so-called "open" macrotexture. Indeed, the inventor noted that for these soil natures, the frequency spectrum was specific, highlighting more easily significant spectral emergences of the various modalities other than those associated with the soil nature class.
[0034] In addition, this specific soil nature class allows the selection of vibroacoustic measurements made on soil with a rather open macrotexture, which makes it possible to limit the number of measurements eligible for the prognosis of the tire wear condition, which makes the method more efficient.
[0035] The specific soil nature class could include soils with a PMT greater than 0.7, preferably greater than 0.9, very preferably greater than 1.0.
[0036] Indeed, the inventor found that the method could be just as effective when using soils whose PMT was less than 1.1, which amounts to so-called "medium" soils. However, only the high PMT range within so-called "medium" soils allows for effective implementation of the method, as when the PMT is greater than 0.7. Of course, the higher the PMT and the closer it is to the limit of so-called "open" soils, the more robust the method.
[0037] According to a third embodiment, the determination of the soil nature class includes the additional steps: Recording a fourth measurement of a vibroacoustic signal produced by the tire rolling on the road surface during a fourth given time frame; Transforming the fourth time signal into a fourth frequency signal over a fourth given frequency interval; Segmenting the fourth frequency interval into at least two frequency bands of predetermined width and associating with each at least two frequency bands at least one data item representative of the fourth frequency signal in said at least one frequency band, the at least one representative data item from the fourth measurement forming at least one variable of a matrix associated with said fourth measurement;Determine the soil type class corresponding to the matrix associated with the fourth measurement carried out, using a fourth automatic data learning based on a learning base formed from a set of matrices associated with measurements recorded and carried out according to the same steps as above, in known rolling conditions according to modalities each representing at least one soil type class.;
[0038] The nature of the road surface can be achieved through a vibroacoustic measurement of the tire rolling on the surface of a road. For this, the use of a fourth automatic learning built on a learning base taking into account as modality the classes of nature of the ground is recommended. Indeed, the vibroacoustic response of the tire depends on the state of wear of the tire but also on other parameters such as the rolling speed, the meteorological state of the ground. However, the rolling speed influences the entire frequency spectrum of the tire and not only certain specific frequency bands or at least these bands are not all identical according to the observed parameters. Thus, the data representative of the vibroacoustic signature matrix linked to the nature of the ground is different from that of the state of wear of the tire or that of the rolling speed or that of the meteorological state of the ground.Therefore, it is possible through specific machine learning to identify the nature of the soil or the soil nature class corresponding to a vibroacoustic measurement of the tire by dividing the frequency signal into specific frequency bands and evaluating particular representative data on each frequency band. This frequency division and its representative data are by nature different from those associated with the tire wear state identification matrix, those associated with the matrix linked to the rolling speed and those associated with the matrix linked to the meteorological state of the soil. However, having previously identified the rolling speed class, it is possible to take this parameter into account to reduce the learning base of this fourth machine learning, which makes the method more efficient.Of course, this fourth vibroacoustic measurement can be different from the vibroacoustic measurement of the state of wear, the vibroacoustic measurement of the rolling speed and the vibroacoustic measurement of the meteorological state of the ground. However, nothing prevents the same vibroacoustic measurement from being used as one or other of the three other vibroacoustic measurements associated respectively with the three other automatic learning processes, which reduces the number of steps to be carried out.
[0039] In this particular embodiment, it is necessary to identify the soil type class before determining the tire wear condition, which requires a certain order of post-processing. On the other hand, it is possible to perform the post-processing of the vibroacoustic measurement associated with the soil type class before or after the post-processing of the vibroacoustic measurement of the soil meteorological condition class. And it may become mandatory to perform the post-processing of the vibroacoustic measurement associated with the soil type class after the post-processing of the vibroacoustic measurement of the rolling speed class to take this parameter into account in the fourth automatic learning.
[0040] Optionally, the soil nature class determination step includes the soil surface condition class as a modality of the fourth machine learning.
[0041] The inventor found that positioning the determination of the meteorological state of the soil before the determination of the nature of the soil makes it possible to take into account the meteorological state of the soil as a modality of the fourth automatic learning associated with the nature of the soil, which makes it possible to improve the prognosis associated with this fourth automatic learning. This makes the determination of the nature of the soil more efficient by reducing the number of vibroacoustic measurements to be analyzed to identify the nature of the soil.
[0042] Advantageously at least one of the machine learning methods is included in the group comprising neural networks, discriminant analysis, support vector machines, boosting, k-nearest neighbor methods and logistic regression.
[0043] Machine learning methods are numerous, and each has its own advantages and disadvantages. The list presented is not exhaustive, and the method should not be limited to these concrete examples of machine learning.
[0044] Advantageously, the transformation of a time signal into a frequency signal is carried out using the Power Spectral Density.
[0045] The inventor found that using Power Spectral Density to transform the time signal into a frequency signal has the well-known advantages of conserving the energy of the signal while reducing the parameters of the frequency signal to a single quantity, the amplitude, eliminating the notion of phase.
[0046] Preferably, the state of wear of the tire is included in the group comprising new and worn, preferably also semi-worn.
[0047] In order to make the method efficient, it is necessary to define at least two tire wear states, which allows the vehicle parameters to be adjusted in a binary mode. Thus, the tire's life is split into two periods: a new state and a worn state, for which the tire's behavior changes from a first state to a second state, which justifies adapting the safety devices, for example, the ABS device (acronym for Anti Blocking System) according to each of these two states. Of course, it is entirely possible to split the tire's life into a larger number of tire states. It is advisable that these states all correspond to the same variation in tread height.However, the more the tire tread is cut, the more cumbersome the method is. The inventor found that the best compromise between performance and efficiency of the method was to split the tire tread into three states: worn, half-worn and new. Brief description of the drawings
[0048] The invention will be better understood on reading the following description, given solely as a non-limiting example and with reference to the appended figures in which the same reference numbers designate identical parts throughout and in which: There figure 1 presents a general overview of the method for estimating the state of wear of a tire according to the invention; The figure 2 This represents a vehicle equipped with a measuring device allowing the determination of a state of wear of the tire; The figure 3represents a discriminating space along two main axes of the method for estimating the state of wear of the tire taking into account the modalities of the state of wear of the tire according to the invention; The figure 4 represents noise spectra measured on a vehicle traveling on a road whose state and nature of the ground surface are known; The Figure 5 represents a discriminating space along a main axis of the method for determining the class of ground surface condition taking into account only the modality linked to the ground surface condition according to the invention; The figure 6 represents a discriminating space along two main axes of the method for estimating the state of wear of the tire taking into account the modalities of the state of wear of the tire and the classes of rolling speed for a specific class of ground condition, the so-called “dry” state, according to the invention; The figure 7represents a discriminating space according to two main axes of the method for determining the soil nature class according to the invention; The figure 8 represents a discriminating space along two main axes of the tire wear condition estimation process by selecting measurements according to the criteria of specific soil condition class and specific soil nature class. Detailed description of embodiments
[0049] There figure 1 presents a complete synopsis of the method for estimating the state of wear of a tire using a vibroacoustic measurement 1001 carried out on a vehicle during driving according to several embodiments.
[0050] This 1001 time measurement is transformed into a 1002 frequency spectrum by standard calculation tools such as the Fast Fourier Transform.
[0051] The frequency spectrum 1002 is then divided into different frequency bands according to the chosen use. On each frequency band, one or more physical quantities of the reduced frequency spectrum on the frequency band are associated with the frequency spectrum. The set of quantities constitutes a vector whose length is proportional to the number of physical quantities calculated. This makes it possible to fill a matrix 1003, one of the dimensions of which is the number of frequency bands obtained from the full width of the frequency spectrum. The second dimension of the matrix 1003 corresponds to the maximum number of physical quantities evaluated per selected frequency band. Generally, the matrix is a vector whose length is the number of selected frequency bands and the second dimension is a scalar. It can also be a two-dimensional matrix or the second dimension is a vector.
[0052] In a first embodiment, the matrix 1003 is injected into a machine learning 1004 comprising a learning base. The learning base was formed during a prior learning step by a succession of vibroacoustic measurements and frequency processing on the time measurement where the modalities of the learning base were mastered. In a conventional mode, the modalities are the tire wear state classes comprising at least the worn state and the new state, preferably also the semi-worn state. The machine learning delivers a prognosis on the tire wear state.
[0053] The repetition of measurements, post-processing and forecasts makes it possible to constitute a series M of forecast results. The repetition of the same result regularly in the series makes it possible to validate the change in the tire's wear state. In addition, generally, the initial wear state of the tire is the new state which transforms over time and therefore from the forecast series to the semi-worn state and then the worn state. Thus, the redundancy of the forecast results on the same wear state, knowing that the change in wear state can generally only be done in one direction, makes it possible to quickly determine the actual wear state of the tire in the form of a wear state class.The higher the number of wear classes, and therefore the greater the precision in determining the state of wear of the tire, the less efficient the method is since the quality of the prognosis result will be affected by all the influential parameters other than the state of wear of the tire. Examples include driving speed, the nature of the ground, the meteorological state of the ground, but also the vehicle, the conditions of use of the tire in terms of pressure, applied load, external temperature, etc.
[0054] According to a second embodiment, in order to robustify the method for estimating the state of wear of the tire, the prognosis by automatic learning 1004 can also take into account the rolling speed of the tire 2001. Indeed, this parameter greatly influences the average level of the frequency spectrum 1002 obtained from the temporal vibroacoustic measurement 1001. Taking this parameter into account in the form of rolling speed classes as a modality of automatic learning makes it possible to reduce erroneous prognoses.
[0055] This rolling speed of the tire can be obtained by additional sensors of the vehicle or obtained through information passing through the electronic wiring of the vehicle. This determination can be made directly in the form of classes or arranged in the form of classes of desired rolling speed by machine learning. However, in a variant, the determination of the rolling speed class 2002 is determined using a second vibroacoustic measurement 1001 obtained on the vehicle. This second vibroacoustic measurement 1001 is advantageously the vibroacoustic measurement 1001 which will be used to predict the state of wear of the tire in step 1004.
[0056] In the same way as previously, the temporal vibroacoustic measurement 1001 is transformed into a frequency spectrum 1002. This frequency spectrum is then divided into frequency bands. One or more physical quantities of the frequency spectrum are associated with each frequency band. This makes it possible to complete a matrix 1003 associated with the prognosis of the rolling speed class. However, the division into frequency bands has no reason to be similar to that carried out for the prognosis 1004 of the tire wear condition. The speed class identification 2002 is largely sufficient for the prognosis 1004 of the tire wear condition.
[0057] Optionally, the meteorological surface condition 3001 of the ground on which the vehicle is traveling can also be determined in order to robustify the process for estimating the state of wear of the tire; the prognosis by automatic learning 1004 can also take this meteorological condition into account in two distinct ways.
[0058] The first way is to determine classes 3002 of ground condition and to take into account these classes of ground condition as a modality of the prognosis 1004 of the state of wear of the tire. Knowing the meteorological surface state of the ground, at least in terms of classes, makes it possible to robustify the prognosis at the expense of a more substantial learning base and a more complex mathematical model.
[0059] The second way consists of determining a specific ground condition class 3003 for which the prognosis of the tire wear condition will be carried out. In fact, we select from the set of vibroacoustic measurements 1001 those which correspond to a ground condition favoring the prognosis 1004 of the tire wear condition. Thus, the learning base associated with the prognosis is reduced and the mathematical model more elementary, which allows fast calculation times with few resources and a robust prognosis.
[0060] The meteorological surface condition 3001 of the ground is divided into various classes. In summer or normal conditions, a distinction is made between dry and wet conditions, and the latter is even separated according to the depth of water on the road. To this can also be added winter conditions such as icy or snowy conditions.
[0061] The applicant noted that the specific soil condition class must include the so-called dry condition. Indeed, this soil condition class has a high occurrence in the 2001 vibroacoustic measurements and allows for greater reproducibility in the frequency band division at step 1003. Thus, focusing on the vibroacoustic measurements where the soil condition class is specific is not detrimental in terms of the occurrence of the 1001 vibroacoustic measurements in a large majority of territories while being effective in terms of prognosis due to the similarity of the 1002 frequency spectra for this meteorological condition.
[0062] Of course, this consideration of ground condition classes can be done with or without taking into account the rolling speed 2001 in the automatic learning 1004. However, the determination of ground condition class 3002 will necessarily be subsequent to the determination of speed class 2002 if we wish to take everything into account.
[0063] Finally, the meteorological condition of the ground 3001 can be obtained by additional sensors of the vehicle such as a rain detector on the windshield, or the actuator of the windshield wiper trigger or obtained through information passing through the electronic wiring of the vehicle. However, in a variant, the determination of the condition class 3002 of the ground is determined using a third vibroacoustic measurement 1001 obtained on the vehicle. This third vibroacoustic measurement 1001 is advantageously the vibroacoustic measurement 1001 which will be used to predict the state of wear of the tire 1004 and / or the vibroacoustic measurement 1004 which was used for the determination 2004 of the rolling speed class 2002.
[0064] In the same way as previously, the temporal vibroacoustic measurement 1001 is transformed into a frequency spectrum 1002. This frequency spectrum is then divided into frequency bands. One or more physical quantities of the frequency spectrum are associated with each frequency band. This makes it possible to complete a matrix 1003 associated with the determination of the meteorological condition class 3002 of the ground. However, the division into frequency bands has no reason to be similar to that carried out for the prognosis 1004 of the tire wear condition or for the determination of the rolling speed class 2002. In fact, the identification of the ground condition class 3002 is largely sufficient at the level of the prognosis 1004 of the tire wear condition.
[0065] Optionally, the nature 4001 of the texture of the ground where the vehicle is driving can also be determined in order to robustify the process for estimating the state of wear of the tire, the prognosis by automatic learning 1004 can also take into account classes of nature of the ground according to two distinct paths.
[0066] The first way is to determine classes 4002 of soil nature and to take into account these classes of soil nature as a modality of the prognosis 1004 of the state of wear of the tire. Knowing the nature of the soil 4001, at least in terms of classes, makes it possible to robustify the prognosis 1004 at the expense of a more substantial learning base and a more complex mathematical model.
[0067] The second way consists of determining a specific soil type class 4003 for which the prognosis 1004 of the tire wear condition will be carried out. In fact, we select from the set of vibroacoustic measurements 1001 those which correspond to a soil type favoring the prognosis 1004 of the tire wear condition. Thus, the learning base associated with the prognosis is reduced and the mathematical model more elementary, which allows rapid calculation times with few resources and a robust prognosis.
[0068] The nature of the soil 4001 of the soil is separated into various classes according to the roughness on a millimeter scale. This nature of the soil is characterized by the PMT.
[0069] The applicant noted that the specific soil type class must include soils qualified as "open." Indeed, this class allows for greater reproducibility in the frequency band division at step 1003. Thus, focusing on vibroacoustic measurements where the soil type class is specific is effective in terms of prognosis due to a certain similarity of the frequency spectra 1002. However, it is entirely possible to extend the specific soil type class to all soils with a PMT greater than 0.7, which also covers the upper part, in terms of PMT, of soils qualified as "medium."
[0070] Of course, this consideration of soil nature classes can be done with or without taking into account the rolling speed 2001 or the surface condition 3001 in the automatic learning 1004. However, the consideration of the soil nature class will necessarily be subsequent to the determination of the speed class 2002 if we wish to take these two parameters into account.
[0071] Finally, the nature of the ground 4001 can be obtained by additional sensors of the vehicle such as laser or sound measuring devices or obtained through information passing through the electronic wiring of the vehicle. However, in a variant, the determination of the ground nature class 4002 is determined using a fourth vibroacoustic measurement 1001 obtained on the vehicle. This fourth vibroacoustic measurement 1001 is advantageously the vibroacoustic measurement 1001 which will be used to predict the state of wear of the tire in step 1004 and / or the vibroacoustic measurement 1001 which was used to determine the rolling speed class 2002 and / or the vibroacoustic measurement 1001 which was used to determine the ground condition class 3002.
[0072] In the same way as previously, the temporal vibroacoustic measurement 1001 is transformed into a frequency spectrum 1002. This frequency spectrum is then divided into frequency bands. One or more physical quantities of the frequency spectrum are associated with each frequency band. This makes it possible to complete a matrix 1003 associated with the determination 4002 of the soil nature class 3002. However, the division into frequency bands has no reason to be similar to that carried out for the prognosis 1004 of the tire wear condition or for the determination of the rolling speed class 2002 or for the determination of the soil condition class 3002. In fact, the identification of the soil nature class 3002 is largely sufficient at the level of the prognosis 1004 of the tire wear condition.
[0073] Finally, the applicant noted that taking into account the meteorological condition and then the nature of the ground makes it possible to robustify the prognosis 1004 of the state of wear of the tire. Indeed, the separating power of the ground condition class is stronger than the separating power of the ground nature class.
[0074] In conclusion, the optional inclusion of soil condition class 3002 and soil nature class 4002 makes it possible to robustify the forecast. The combination is the most effective configuration in the above-mentioned order. The effectiveness is measured by the forecast error rate.
[0075] However, by taking the specific class route, the method gains in efficiency through the reduced size of the learning base of the prognosis 1004 of the state of wear of the tire and the speed and simplicity of the calculations which can be carried out on board and in real time on the vehicle.
[0076] In the main embodiment, the sound signal generated by the tire (T) is measured using a microphone (1) placed in the vehicle. On the Figure 2 , the microphone is placed in the front part of a wheel arch located at the rear of the vehicle (C). However, other locations can be considered, such as, for example, the rear bumper. The choice of location for the microphone depends, for example, on the type of data to be estimated, the type of vehicle and the external constraints related to its installation, maintenance, and durability.
[0077] On the Figure 2 , we visualize a vehicle C rolling on ground G, represented schematically, comprising front and rear wheel arches in which the wheels equipped with tires T are housed.
[0078] When vehicle C moves, tire T generates a noise whose amplitude and frequency depend on multiple factors. This sound pressure is in fact the superposition of noises of various origins such as the noises generated by the contact of the tread blocks with the ground G, by the movements of air between the tread elements, by the water particles raised by the tire, or by the air flows linked to the speed of the vehicle. Listening to these noises also overlaps with noises linked to the vehicle's environment such as the noise of the engine. All these noises are also dependent on the speed of the vehicle.
[0079] A measuring means, such as a microphone 1, is installed in the vehicle. It is specified here that different positions can be envisaged for the measuring means, only one being shown on the figure 1, but not limiting the scope of the invention to this configuration. Thus, the sensor can be positioned on a wall of the rear bumper, for example, but not necessarily oriented so as to detect an audible signal coming from the rear of the vehicle.
[0080] Another option is to position the microphone on the front bumper of a vehicle. The measuring device can also be positioned in a wheel arch to listen to rolling noise as close as possible to where it is generated. Ideally, installing a vibroacoustic sensor in each wheel arch is the best way to capture all rolling noise and vibrations generated by the tires. However, to determine the condition of the ground (weather conditions) and the nature of the ground (macro-texture of the road surface), a single microphone is sufficient. In the latter case, it is preferable to isolate it from aerodynamic and engine noise.
[0081] Of course, the usual precautions are taken to protect the measuring device from external aggressions such as splashes of water, mud or gravel.
[0082] The vehicle also comprises a computer 2, connected to the measuring means, and configured to execute the operations making it possible to format and analyze, as will be described in detail later, the raw information coming from the measuring means, and to estimate the state of the tire based on a measurement of the vibroacoustic emission detected by the measuring means.
[0083] There figure 3is a visualization according to the state of wear of the tire divided into three classes of identical size for vehicle rolling speeds between 20 and 130 km / h. The first class is the so-called "new" wear class corresponding to an upper third of the useful height of the tire tread. The useful height is delimited by the maximum height of the tread corresponding to the radially outer end of the tire in relation to its natural axis of rotation and the minimum height is defined by the radially outer end of the wear indicator at the bottom of the groove. The second class called "half-worn" corresponds to the middle third of the useful height of the tread. Finally, the last class called "worn" corresponds to the lower third of the useful height of the tread.
[0084] Here we visualize a set of vibroacoustic measurements carried out on the vehicle regardless of the vehicle conditions in terms of applied load and inflation pressure starting from the nominal configuration on the vehicle. The vehicle completed a road circuit including an urban route, a route on a country road and a motorway route over several sessions in the intermediate season, which makes it possible to mix all the conditions on the ground surface condition, in particular the so-called "dry", "humid" and "wet" ground conditions and on various types of ground surface nature, in particular so-called "closed", "medium" and "open" soils. The learning base takes into account as a modality only the state of wear of the tire. The measurements are made here with tires which have the three types of tire wear condition.
[0085] Machine learning in relation to its conditions defines main directions mathematically. We visualize on the figure 3, the discriminant space according to the first two principal directions which represent the two axes of representation of the graph. Machine learning identifies three series of circles in this two-dimensional representation. The first series, represented by dotted circles, represents the various probabilities of the so-called "new" state of wear in this two-dimensional discriminant space. The circles are concentric. The highest probability, i.e. greater than 0.9, is defined by the area delimited by the smallest circle. Then, the next circle defines a decreasing probability of 0.1, i.e. a value of 0.8. And, the following circle represents a probability having further decreased by 0.1, i.e. the value of 0.7. The second series of circles represented by circles in solid gray lines represents the various probabilities of the so-called "half-worn" state of wear in this two-dimensional discriminant space.Finally, the third series of circles represented by black solid line circles represents in the same way the various probabilities of the state of wear called "worn". We note that the three series of circles are globally disjoint, in particular at the level of their small circle called main circle which allows to sort the measurements according to the three states of wear. However, the circles are voluminous and the secondary circles have overlaps between them. As a result, there is uncertainty on the forecast made. It is possible that a wrong forecast is made. Thus, by multiplying the measurements, statistically we minimize the wrong forecasts in all the forecasts which allows to determine the state of wear of the tire. In this configuration, it is necessary that the series of forecasts N having identified the same state of wear is consistent in a series M of important forecasts. Here, we assign to each measurement a point symbol.The round symbol "o" represents a measurement whose prognosis indicates a tire with intermediate wear. The plus symbol "+" represents a measurement whose prognosis indicates a tire with severe wear. Finally, the cross symbol "x" represents a measurement whose prognosis indicates a tire in near-new condition.
[0086] There figure 4 is a spectral representation of the sound power recorded by the microphone during a time frame. Here, a time frame is understood to be the time interval, generally short, during which a recording is made on the basis of which the data used as the basis for a measurement are established. This time frame is less than or equal to 0.5 seconds or ideally less than or equal to 0.25 seconds.
[0087] This spectral representation represents the received sound power (in dB) as a function of frequency, over a given frequency interval, typically here, the audible frequency interval, between 0 Hz and 20 KHz.
[0088] More specifically the spectral representation of the figure 4 is obtained by decomposing the frequency interval into frequency bands of predetermined widths, and assigning to each frequency band a characteristic value equal to the average power measured in this frequency band. Here, a division of the frequency interval into third-octave bands has been used. Thus, each point of each of the curves of the figure 4 represents an average sound power for a given frequency band and measured during a time frame in driving conditions in which, all other things being equal, only the driving speed is varied (typically from 30kmh to 110kmh).
[0089] It is then observed that the curves representing the spectral powers are offset from each other, and that the total dissipated sound power increases as a function of speed. Although the general shape of the curves remains similar, which justifies that this parameter does not systematically upset the prognosis of the state of wear, certain specificities of the spectrum are more or less marked which generates noise in the prognosis constructed on its spectra. Taking into account the driving speed, in particular in terms of a speed class of approximately 30 km / h, makes it possible to improve the prognosis of the state of wear of the tire according to a second embodiment of the invention.
[0090] These observations are reproduced when one or more modalities of the other classes are changed and the curves obtained by varying only the speed parameter are compared.
[0091] There Figure 5is a two-dimensional representation of a series of measurements taken on a vehicle for the same set of tires, regardless of its state of wear and regardless of its conditions of use on the vehicle during a journey on various roads. However, some measurements were taken on damp or wet ground, as evidenced by the activation of the vehicle's windshield wipers. However, since no measurement of the height of water on the roads was recorded, the classification of these measurements is made according to two unique classes: dry and damp / wet.
[0092] There Figure 5highlights the effectiveness of the third automatic learning of the process for determining the state of the ground surface, in the form of a class, from a vibroacoustic temporal measurement. In a discriminant space, here in a representation through the single principal vector, the measurements are easily classified according to the two aforementioned modalities. The graph represents for a set of measurements indexed from 1 to 1290 along the first axis of the graph and arranged according to the wiper activation indicator, the norm of the principal vector along the vertical axis of the graph. We can easily distinguish the two classes of ground state, forming two groups of measurements, whose values of the norm of the principal vector are strongly differentiated. In addition, we find a notion of Gaussian distribution of the measurements around a central value for each group.Thus, the discriminating power of the vibroacoustic measurement according to the third machine learning is verified on a principal vector from the vibroacoustic measurement on the vehicle. Although some points are aberrant, their limited number easily leads to determining the meteorological state of the ground where the vehicle is traveling with low measurement redundancy. The other solution consists of not taking into account these adherent points, since an uncertainty remains, and to reinitialize a process for estimating the state of wear of the tire.
[0093] There figure 6 is a visualization according to the state of wear of the tire divided into three classes of identical size, the so-called "new" class, the so-called "half-worn" class and the "worn" class, for vehicle driving speeds between 20 and 130 km / h.
[0094] Here we visualize a set of vibroacoustic measurements carried out on the vehicle regardless of the vehicle conditions in terms of applied load and inflation pressure starting from the nominal configuration on the vehicle. The vehicle completed a road circuit including an urban route, a route on a country road and a motorway route over several sessions in the intermediate season, which makes it possible to mix all the conditions on the ground surface condition, in particular the so-called "dry", "humid" and "wet" ground conditions and on various types of ground surface nature, in particular so-called "closed", "medium" and "open" grounds. The learning base takes into account as a modality the state of wear of the tire but also the vehicle's rolling speed class. The measurements are made here with tires which have the three types of tire wear condition.However, a condition for carrying out the prognosis of the tire wear condition is set up on the ground surface condition through the specific ground condition class corresponding to the so-called "dry" ground. Thus, if the ground surface condition does not correspond to the specific ground condition class, the prognosis on the tire wear condition by machine learning is not carried out. This excludes a number of vibroacoustic measurements but the occurrence of the measurements on the vehicle is sufficient compared to the slow temporal evolution of tire wear.
[0095] Machine learning in relation to its conditions defines main directions mathematically. We visualize on the figure 6, the discriminant space according to the first two principal directions which represent the two axes of representation of the graph. Machine learning identifies three series of circles in this two-dimensional representation. The first series, represented by dotted circles, represents the various probabilities of the so-called "new" state of wear. The circles are concentric with each other in this two-dimensional discriminant space. The highest probability, i.e. greater than 0.9, is defined by the area delimited by the smallest circle. Then, the next circle defines a decreasing probability of 0.1, i.e. a value of 0.8. And, the following circle represents a probability having further decreased by 0.1, i.e. the value of 0.7. The second series of circles represented by circles in solid gray lines represents the various probabilities of the so-called "half-worn" state of wear in this two-dimensional discriminant space.Finally, the third series of circles represented by black continuous line circles represents in the same way the various probabilities of the state of wear called "worn". We note that the three series of circles are globally disjoint, more so than those of the . figure 3, especially at the level of their smallest circle called the main circle, which allows to sort the measurements more efficiently according to the three states of wear. However, the circles are large and the secondary circles have overlaps between them. As a result, there is uncertainty about the forecast made. It is possible that a wrong forecast is made. Thus, by multiplying the measurements, statistically we minimize the wrong forecasts in all the forecasts which allows to determine the state of wear of the tire. Here, the fact of discarding a certain number of vobroacoustic measurements because of the class of ground surface state, robustifies the method by minimizing the wrong forecasts. In this configuration, it is necessary that the series of forecasts N having identified the same state of wear is consistent in a series M of important forecasts. Here, we assign to each measurement a point symbol.The round symbol "o" represents a measurement whose prognosis indicates a tire with intermediate wear. The plus symbol "+" represents a measurement whose prognosis indicates a tire with severe wear. Finally, the cross symbol "x" represents a measurement whose prognosis indicates a tire in almost new condition.
[0096] There figure 7is a two-dimensional representation of a series of measurements made on vehicles for similar sets of tires, regardless of their state of wear and regardless of their conditions of use on vehicles during journeys on various roads, and regardless of weather conditions. The roads used are vehicle test tracks whose surface texture is regularly checked by a PMT-type measurement. The fourth machine learning, taking into account that the modalities associated with the three soil nature classes, "open", "medium" and "closed", defines principal directions characterized by vectors. Here, the discriminant space is represented using the two principal vectors to dimensionally represent the discriminant space.
[0097] The fourth machine learning identifies three series of circles in this two-dimensional representation related to each of the modalities. The first series, represented by dotted circles, represents the various probabilities of the nature of the so-called "medium" soil. The circles are concentric with each other in this two-dimensional discriminant space. The highest probability, i.e. greater than 0.9, is defined by the area delimited by the smallest circle. Then, the next circle defines a decreasing probability of 0.1, i.e. a value of 0.8. And, the following circle represents a probability that has further decreased by 0.1, i.e. the value of 0.7. The second series of circles, represented by circles in solid gray lines, represents the various probabilities of the nature of the so-called "closed or smooth" soil in this two-dimensional discriminant space.Finally, the third series of circles represented by black solid line circles represents in the same way the various probabilities of the nature of the soil called "open or macrorough". We note that the three series of circles are globally disjoint, perfectly between the classes called "open" and "closed", and at least at the level of their smallest circle called main circle which allows to sort the measurements according to the three natures of soil. However, there remains an uncertainty on the prognosis made on certain points. It is possible that a wrong prognosis is made. Thus, by multiplying the measurements, statistically we minimize the wrong prognoses in the totality of the prognoses which allows to determine the nature of the soil according to the three classes mentioned above. Here, the fact of discarding a certain number of vibroacoustic measurements because of the class of state of the soil, would robustify the method by minimizing the wrong prognoses.In this configuration, it is necessary that the series of forecasts N having identified the same state of wear be consistent in a series M of important forecasts. Here, each measurement is assigned a point symbol. The round symbol "o" represents a measurement whose forecast indicates a soil whose nature would be of the medium type. The plus symbol "+" represents a measurement whose forecast indicates a smooth or closed type of soil. Finally, the cross symbol "x" represents a measurement whose forecast indicates a macrorugious or open type of soil.
[0098] There figure 7highlights the effectiveness of the fourth machine learning of the method for determining the nature of the soil from a vibroacoustic temporal measurement. In the discriminating space, here in a representation taking into account the two main directions, the measurements are easily classified according to the three aforementioned modalities. The three classes of soil nature are easily identified by machine learning and distinguished, forming three groups of measurements, whose values of the norm of the main vectors are strongly differentiated, particularly for the first vector. Thus, the discriminating power of the vibroacoustic measurement according to the fourth machine learning is validated by this two-dimensional representation resulting from the vibroacoustic measurement on a vehicle. Although some points are aberrant, their limited number easily leads to determining the texture of the soil where the vehicle is traveling with low redundancy of measurements.
[0099] There figure 8 is a visualization according to the state of wear of the tire divided into three classes of identical size, the so-called "new" class, the so-called "half-worn" class and the "worn" class, for vehicle driving speeds between 20 and 130 km / h.
[0100] Here we visualize a set of vibroacoustic measurements carried out on the vehicle regardless of the vehicle conditions in terms of applied load and inflation pressure starting from the nominal configuration on the vehicle. The vehicle completed a road circuit including an urban route, a route on a national road and a motorway route over several sessions in the intermediate season, which makes it possible to mix all the conditions on the ground surface condition, in particular the so-called "dry", "humid" and "wet" ground conditions and on various types of ground surface nature, in particular so-called "closed", "medium" and "open" grounds. The learning base takes into account as a modality the state of wear of the tire but also the rolling speed class of the tire. The measurements are made here with tires which have the three types of tire wear condition.However, two conditions for performing the tire wear condition prognosis are required before undertaking the prognosis, one on the ground surface condition and the other on the nature of the ground. Thus, if the ground surface condition does not correspond to the specific ground condition class, the tire wear condition prognosis by machine learning is not performed. Similarly, if the ground nature does not correspond to the specific ground nature class, whose PMT is greater than 0.7, the tire wear condition prognosis by machine learning is not performed. This excludes a number of vibroacoustic measurements but the occurrence of the measurements on the vehicle is sufficient compared to the slow temporal evolution of tire wear.
[0101] Machine learning in relation to its conditions defines main directions mathematically. We visualize on the figure 8, the discriminant space according to the first two principal directions which represent the two axes of representation of the graph. Machine learning identifies three series of circles in this two-dimensional representation. The first series, represented by dotted circles represents the various probabilities of the so-called "new" state of wear. The circles are concentric with each other in this two-dimensional discriminant space. The highest probability, that is to say greater than 0.9, is defined by the interior surface delimited by the smallest circle. Then, the next circle defines a decreasing probability of 0.1, or a value of 0.8. And, the following circle represents a probability having decreased again by 0.1, or the probability value of 0.7.The second series of circles represented by gray solid line circles represents the various probabilities of the so-called "half-worn" wear state in this two-dimensional discriminant space. Finally, the third series of circles represented by black solid line circles represents in the same way the various probabilities of the so-called "worn" wear state. We note that the three series of circles are globally disjoint, more so than those of the . figures 3 And figure 6, particularly at the level of their smallest circle called the main circle, which makes it easy to sort the measurements according to the three wear states. As a result, there is little uncertainty about the forecast made. It is possible that an incorrect forecast is made. Thus, in a few measurements, statistically we minimize the incorrect forecasts in all the forecasts, which makes it possible to determine the tire wear state. Here, the fact of excluding a certain number of vibroacoustic measurements due to the soil condition class and the soil nature class robustifies the method by reducing the incorrect forecasts. In this configuration, it is necessary that the series of forecasts N having identified the same wear state be initiated in a series M of consecutive forecasts to decide on the wear state. Here, each measurement is assigned a point symbol. The round symbol "o" represents a measurement whose forecast indicates a tire with intermediate wear.The plus symbol "+" represents a measurement whose prognosis indicates a tire with severe wear. Finally, the cross symbol "x" represents a measurement whose prognosis indicates a tire in almost new condition. No erroneous sorting in this example is identified. Finally, the fact of having chosen so-called specific classes in terms of conditions for carrying out the prognosis makes it possible to limit the calculation times and the calculation resources required with automatic learning taking into account all of these classes as a modality of automatic learning. This amounts to minimizing the learning base to only the modalities retained by automatic learning. In addition, this justifies at a lower cost an exploitation of the process for estimating the state of wear of the tire on board and in real time on the vehicle.
[0102] The applicant found that taking into account a specific rolling speed class also improves the quality of the prognosis of the tire wear condition. However, the gain in prognosis quality is offset by the occurrence of vibroacoustic measurements unless a broad specific rolling speed class is defined.
[0103] Of course, between the method revealed at the figure 3 and the method revealed at the figure 8 , there are several possible variants concerning the role of a class by alternating between the choice between a modality of automatic learning or a condition for carrying out automatic prognosis by defining specific classes.
Claims
1. A method for estimating the state of wear of a tire of a mounted assembly of a vehicle traveling on a surface of a road, comprising the following steps: - measuring a vibroacoustic signal (1001) produced by the tire running on the road surface during a given time frame; - converting the time signal (1001) into a frequency signal (1002) over a given frequency range; - segmenting the frequency range into at least one frequency band with a predetermined width and associating at least one datum representing the frequency signal in said at least one frequency band with the at least one frequency band, with the at least one representative datum derived from the measurement forming at least one variable of a matrix (1003) associated with said measurement; - predicting a state of wear of the tire corresponding to the matrix associated with the completed measurement, by means of machine learning (1004) from the data based on a learning database made up of a set of matrices associated with measurements recorded and carried out according to the same steps as above, under known running conditions, according to modalities each representing a state of wear of the tire; - determining a ground surface state category (3002); and wherein the ground surface state category (3002) is a condition for predicting the state of wear of the tire by the machine learning step (1004) if it is a specific ground state category (3003) containing the "dry" ground surface state category or the "dry" and "wet" ground surface state categories; - determining a ground nature category (4002); and wherein the ground nature category (4002) is a condition for predicting the state of wear of the tire by the machine learning (1004) if it is a specific ground nature category (4003) comprising the ground nature category referred to as "open"; and - determining the state of wear of the tire (1005) after a number N of identical predictions in a series M of consecutive predictions.
2. The method for estimating the state of wear of a tire according to claim 1, wherein the method comprises the following step: - determining a tire running speed category (2002), the breadth of which is a fraction of a maximum running speed, preferably the maximum running speed is 300 km / h; and wherein the running speed category (2002) is a modality of the machine learning (1004).
3. The method for estimating the state of wear of a tire according to the preceding claim, wherein determining the running speed category (2002) of the tire comprises the following steps: - recording a second measurement of a vibroacoustic signal (1001) produced by the tire running on the road surface during a second given time frame; - converting the second time signal (1001) into a second frequency signal (1002) over a second given frequency range; - segmenting the second frequency range into at least two frequency bands with a predetermined width and associating at least one datum representing the second frequency signal in said each of the at least two frequency bands with each of the at least two frequency bands, with the at least one representative datum derived from the second measurement forming the at least one variable of a matrix (1003) associated with said second measurement; - determining a running speed category (2002) of the tire corresponding to the matrix associated with the second completed measurement, by means of second machine learning (2004) from the data based on a learning database made up of a set of matrices associated with measurements recorded and carried out according to the same steps as above, under known running conditions, according to modalities each representing a running speed category of the tire.
4. The method for estimating the state of wear of a tire according to any one of claims 1 to 3, wherein determining the ground surface state category (3002) comprises the following additional steps: - recording a third measurement of a vibroacoustic signal (1001) produced by the tire running on the road surface during a given third time frame; - converting the third time signal into a third frequency signal (1002) over a third given frequency range; - segmenting the third frequency range into at least two frequency bands with a predetermined width and associating at least one datum representing the third frequency signal in said each of the at least two frequency bands with each of the at least two frequency bands, with the at least one representative datum derived from the third measurement forming the at least one variable of a matrix (1003) associated with said third measurement; - determining a ground surface state category (3002) corresponding to the matrix associated with the third completed measurement, by means of third machine learning (3004) from the data based on a learning database made up of a set of matrices associated with measurements recorded and carried out according to the same steps as above, under known running conditions, according to modalities each representing at least one ground surface state category.
5. The method for estimating the state of wear of a tire according to any one of claims 1 to 4, wherein determining the ground nature category (4002) comprises the following additional steps: - recording a fourth measurement of a vibroacoustic signal (1001) produced by the tire running on the road surface during a given fourth time frame; - converting the fourth time signal (1001) into a fourth frequency signal (1002) over a given fourth frequency range; - segmenting the fourth frequency range into at least two frequency bands with a predetermined width and associating at least one datum representing the fourth frequency signal in said each of the at least one frequency band with each of the at least one frequency band, with the at least one representative datum derived from the fourth measurement forming the at least one variable of a matrix (1003) associated with said fourth measurement; - determining the ground nature category (4002) corresponding to the matrix associated with the fourth completed measurement, by means of fourth machine learning (4004) from the data based on a learning database made up of a set of matrices associated with measurements recorded and carried out according to the same steps as above, under known driving conditions, according to modalities each representing at least one ground nature category.
6. The method for estimating the state of wear of a tire according to the preceding claim in combination with any one of claims 1 to 5, wherein the step of determining the ground nature category comprises the ground surface state category (3002) as a modality of the fourth machine learning.
7. The method for estimating the state of wear of a tire according to any one of the preceding claims, wherein at least one of the machine learning methods (1004, 2004, 3004, 4004) is included in the group comprising neural networks, discriminant analysis, support vector machines, boosting, K-nearest neighbors methods and logistic regression.
8. The method for estimating the state of wear of a tire according to any one of the preceding claims, wherein the state of wear of the tire is included in the group comprising new and worn, preferably also half-worn, tires.
