Method for determining fluid flow rate in a vehicle propulsion system
A supervised learning neural network-based method estimates fluid flow rates in vehicle engine systems by calculating a hydraulic loss coefficient, addressing the inefficiencies of traditional calibration methods and ensuring accurate flow rate determination with minimal resource usage and maintenance alerts.
Patent Information
- Application Number
- FR2022009609
- Authority / Receiving Office
- FR · FR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2022-09-22
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2042-09-22
AI Technical Summary
Existing methods for determining fluid flow rates in vehicle engine systems, particularly in systems with electronically controlled injectors prone to clogging or wear, rely on tedious calibration table construction and are not robust to changes in passage section due to deposits or wear, especially in open-loop control scenarios.
A method utilizing a loss estimation module, such as a supervised learning neural network, to calculate a hydraulic loss coefficient based on measured parameters, allowing for the estimation of actual fluid flow rates by correcting theoretical calculations, and incorporating an alert system for potential mechanical defects.
Provides accurate estimation of fluid flow rates in vehicle engine systems, even with changes in passage section due to clogging or wear, while requiring minimal computing and memory resources, and offering alerts for maintenance needs.
Smart Images

Figure 00000017_0000 
Figure 00000017_0001 
Figure 00000017_0002
Abstract
Description
Title of the invention: Method for determining fluid flow rate in a vehicle propulsion system Technical field and context
[0001] The present invention relates to methods for determining fluid flow, particularly liquid flow, in a vehicle powertrain system, especially when said flow is not measured directly by a sensor. In this case, it is necessary to estimate said flow (or a quantity delivered) from other data sources and / or other parameters. The present invention is particularly useful when the system includes an electronically controlled injector whose orifice may be subject to deposits or fouling and a reduction in the effective flow area, or which undergoes wear (increasing flow area). Previous art
[0002] It is known, using at least one pressure sensor, to monitor the temporal evolution of the pressure during an injector command sequence. This allows for the recording of the pressure temporal characteristics in the line, as well as the smoothed values of the pressure and other parameters such as temperature. Furthermore, during system development, calibration tables are constructed based on the injector's effective flow area and the various parameters mentioned above.
[0003] The use of these calibration tables and a classic algorithmic logic makes it possible to estimate the actual flow delivered by an injector, whose characteristics can evolve over time, in particular the passage section can decrease due to progressive fouling, or increase again if a crystallized deposit disappears.
[0004] Among the systems of interest here are "selective catalytic reduction" (SCR) systems, which reduce nitrogen oxide emissions. However, the invention can be applied to other liquid injection systems, particularly fuel injection systems, when the control is open-loop. For example, the invention can be applied to a gasoline injection system in which the fuel mixture control loop is not operating momentarily.
[0005] The inventors sought to improve the situation by getting rid of the tedious step of constructing calibration tables. Summary of the invention
[0006] To this end, according to the present disclosure, a method is proposed for determining at least one flow rate value of a fluid of interest in a drive system of a vehicle of interest, said flow rate value not being measured directly by a sensor, the drive system comprising at least one fluid reservoir, a pump, a fluid injection device, with a fluid flow path from the pump to an injection zone downstream of the injection device, and an electronic control unit capable of controlling an opening of the injection device which is otherwise closed in the absence of a command, the fluid of interest advantageously being a liquid fluid which can be considered incompressible, the process comprising the following steps:
[0007] - to have a loss estimation module, providing as output a co efficient hydraulic loss CP
[0008] / b / - to carry out a plurality of N fluid injection sequences, during which one collection of values of a plurality of parameters (dP,Pl,P0,T,X),
[0009] / c / - calculate a theoretical quantity of fluid injected during these N sequences fluid injection, using at least some of the said values of the plurality of parameters (Pl,P0,T,X),
[0010] / d / - transmit, to the loss estimation module, said values of the plurality of parameters (dP,Pl,P0,T,X), and obtain, at the output of the loss estimation module, the loss coefficient CP,
[0011] / e / - calculate an estimated actual quantity of fluid injected during the N sequences fluid injection, by applying the loss coefficient CP to the calculation of the theoretical quantity of fluid.
[0012] Thanks to these provisions, the calculated theoretical quantity can be corrected using the output of the loss estimation module to determine as closely as possible the quantity actually injected. As will be seen later, the loss estimation module can be a supervised learning loss estimation module, for example, of the neural network type. The training could have been carried out beforehand, prior to the actual use by the method described above.
[0013] It should be noted that steps c / and d / can be implemented in parallel. Regarding parameter values, in practice the averaged values over the plurality of N sequences (or the median) are used; the process is thus robust to one or two isolated erroneous values.
[0014] According to one aspect, the loss estimation module can be a supervised learning loss estimation module, the process then comprising a preliminary step:
[0015] / a / - to perform prior training on the loss estimation module at ap supervised testing (52) by means of a series of test injection devices having known flow cross-section characteristics, which are placed successively as injection devices on a similar flow path of a test vehicle to simulate the hydraulic loss introduced by the injection device on the flow path in the vehicle of interest as a function of the plurality of parameters, and by measuring during a sequence of opening of the test injection device, the values of parameters of the plurality of parameters, the loss estimation module taking as input said plurality of parameters (dP,Pl,P0,T,X) and providing as output the hydraulic loss coefficient CP.
[0016] It should be noted that the learning phase (step / a / ) is not necessarily carried out on the same vehicle as the subsequent steps ( / b / to / e / ). In other words, in practice, the learning phase is typically performed on a specific test vehicle where these learning tests can be implemented, and steps / b / and subsequent steps will be carried out on production vehicles manufactured and sold in large numbers and used by a large number of drivers under varying conditions and for significant periods of use, including effects related to wear and aging. It should also be noted that the test injection components may have a smaller flow area than the nominal flow area, or conversely, a larger flow area than the nominal flow area.
[0017] According to one aspect, the theoretical calculation uses a so-called Bernoulli model for an incompressible fluid.
[0018] For example, one can choose a formulation of the type below expressed
[0019] [Math.l] QTH = N Ti ^p(Pl - PO) A,
[0020] Where p is the density or mass density of the fluid, PI is the pressure at the pump outlet, PO is the pressure in the injected zone 4, A is a characteristic section of the passage section, Ti is the injection time, N is the number of injections.
[0021] According to one aspect, it may be provided that in step / e / the loss coefficient CP is applied by multiplying it to the calculation of the theoretical quantity of fluid to obtain the estimated actual quantity of fluid injected during the N fluid injection sequences.
[0022] However, instead of a simple multiplication, certain non-linear effects can be taken into account by applying a non-linear function of the type QRE = F(QTH, CP).
[0023] According to one aspect, in which the loss coefficient CP is between 0 and 1. This corresponds to the case where the section is reduced and the actual quantity is less than the theoretical quantity QTH.
[0024] However, the invention also takes into account the opposite case where the actual quantity is greater than the theoretical quantity, i.e. the cases where CP is greater than 1. In a logic of generalization of the invention, the loss coefficient can be considered as a correction coefficient (e.g. characterizing either a loss or an excess).
[0025] According to one aspect, an alert is activated if the loss coefficient CP is less than a predetermined threshold CPS1, which is preferably between 0.5 and 0.75. A deposit that substantially obstructs the orifice or orifices may require a system check in a garage or even the replacement of a part or component.
[0026] According to one aspect, an alert is activated if the loss coefficient CP is greater than a second predetermined threshold, which is preferably between 1.1 and 1.2. This may denote a mechanical fault which may require repair.
[0027] In one aspect, the loss estimation module comprises a neural network. The neural network may be smaller than 5 kilobytes and preferably smaller than 3 kilobytes. In another aspect, the loss estimation module comprises a neural network of one hundred to several hundred neurons. For example, the neural network in question may comprise fewer than 500 neurons, or preferably fewer than 400 neurons. The memory used is thus small.
[0028] Depending on one aspect, the values of the plurality of parameters (dP,Pl,P0,TKi,X) are filtered and / or smoothed over the N injection sequences for use in the loss estimation module. For each parameter, the average or median of the observed values can thus be calculated. This provides robustness to the process against potential isolated outliers.
[0029] According to one aspect, the plurality of parameters (dP,Pl,P0,TK,X) includes a first parameter (dP) representative of a pressure increase at the closing of the injector ('water hammer' in hydraulic jargon).
[0030] The invention also relates to a system for determining at least one flow rate value of a fluid of interest in a vehicle's drive system, said flow rate value not being measured directly by a sensor, the drive system comprising at least one fluid reservoir, a pump, a fluid injection device, with a fluid flow path from the pump to an injected zone downstream of the injection device, and an electronic control unit capable of controlling an opening of the injection device which is otherwise closed in the absence of a command, the fluid of interest being an incompressible or low-compressibility liquid fluid,
[0031] the system comprising a hydraulic loss estimation module providing as output a hydraulic loss coefficient CP,
[0032] the electronic control unit being configured for:
[0033] / b / - to carry out a plurality of N fluid injection sequences, during which one collection of values of a plurality of parameters (dP,Pl,P0,T,X),
[0034] / c / - calculate a theoretical quantity of fluid injected during these N sequences fluid injection, using at least some of the said values of the plurality of parameters (Pl,P0,T,X),
[0035] / d / - transmit, to the loss estimation module, said values of the plurality of parameters (dP,Pl,PO,T,X), and obtain, at the output of the loss estimation module, the loss coefficient CP,
[0036] ld- calculate an estimated actual quantity of fluid injected during the N fluid injection sequences, by applying the loss coefficient CP to the calculation of the theoretical quantity of fluid.
[0037] Thanks to these provisions, for both the system and the process, the calculated theoretical quantity can be corrected using the output of the loss estimation module to determine as closely as possible the quantity actually injected. This training could be carried out beforehand, prior to the actual use by the system described above.
[0038] It should be noted that steps d and d / can be implemented in parallel. Regarding parameter values, in practice the averaged values over the plurality of N sequences (or the median) are used; the process is thus robust to one or two isolated erroneous values.
[0039] According to one aspect, the loss estimation module may be a supervised learning loss estimation module, the process then comprising a preliminary step:
[0040] / a / - to perform prior training on the loss estimation module at ap supervised testing by means of a series of test injection devices having known passage section characteristics, which are placed in turn as injection devices on a similar flow path of a test vehicle to simulate the hydraulic loss introduced by the injection device on the flow path in the vehicle of interest as a function of the plurality of parameters, and by measuring during an opening sequence of the test injection device, the values of the parameters of the plurality of parameters, the loss estimation module taking as input said plurality of parameters (dP,Pl,P0,TK,X) and providing as output the hydraulic loss coefficient CP.
[0041] The learning conditions are identical or similar to those described above.
[0042] According to one aspect, the fluid can be a urea-based liquid intended to reduce nitrogen oxides. The invention is thus applicable to "selective catalytic reduction" (SCR) systems that reduce nitrogen oxide emissions.
[0043] According to another possibility, the fluid may be a fuel or hydrocarbon intended to be burned in an internal combustion engine.
[0044] In one aspect, the injection device is a needle injector. Such an injector can deliver a small quantity of liquid with high precision. This needle injector has a coil that creates an electromagnetic force when an electric current generated by an electronic control unit passes through it. This provides flexible control that can be adapted to various operating conditions.
[0045] In one aspect, the loss estimation module can be implemented as a neural network within the electronic control unit. This solution offers good availability and is advantageous because it requires little computing power and memory.
[0046] Other aspects, objects, and advantages of the invention will become apparent from the following description of an embodiment of the invention, given by way of non-limiting example. The invention will also be better understood with reference to the accompanying drawings in which:
[0047] - Fig. 1 schematically illustrates a liquid injection system, let's say... works with a controlled injector,
[0048] - Figure 2 schematically represents an injector with nominal characteristics,
[0049] - Figure 3 schematically represents an injector hole with the presence of deposits reducing the passage cross-section
[0050] - Figure 4 schematically represents a cross-section of an injection organ,
[0051] - [Fig. 5] represents a functional block diagram of the system,
[0052] - Figure 6 shows an illustration of the steps in the process,
[0053] - Figure 7 schematically represents an example of a neural network with ap supervised learning,
[0054] - [Fig. 8] represents a pressure timing diagram illustrating a pressure jump observed at the time of closing the injection unit. Detailed description of implementation methods
[0055] In the various figures, the same reference numerals designate identical or similar elements. For the sake of clarity, some elements are not necessarily shown to scale.
[0056] We are interested in systems and methods for determining fluid flow, particularly liquid flow, in a vehicle propulsion system, especially when said flow is not measured directly by a sensor. The vehicle in question may be a car, a truck, or a scooter; the invention can be applied without limitation as to the type of vehicle or its type of propulsion, for example, on a motorboat.
[0057] The flow rate and / or quantity of liquid fluid delivered by an injection device must be estimated from other information intakes and / or other parameters.
[0058] The injection element in the present invention is an electronically controlled injector. The injection element can be placed in an open state when energized by an electrical command provided by an electronic control unit. Conversely, in the absence of an electrical command, the injection element is closed.
[0059] For example, with reference to [Fig. 4], a needle 15 can be provided that is movable relative to an orifice seat 14. In the absence of electrical control, the needle When the needle is positioned on the seat, it closes the orifice (or orifices) and the injection unit is closed. When electrically controlled, the needle is lifted from the seat, thus opening a passage through the orifice (or orifices) and the injection unit is said to be open.
[0060] The orifice (or orifices) 16 of the injection member or the aforementioned seat may be subject to fouling phenomena or solid deposits 18 reducing the passage area, see [Fig.3].
[0061] Conversely, in some cases erosion of the orifice walls (or orifices) can be observed, increasing the passage cross-section. This erosion may be due to cavitation or alteration of the material.
[0062] In the illustrated example, the fluid of interest is a liquid. The liquid is considered incompressible here. Generally, liquids may exhibit slight compressibility.
[0063] According to one example, the fluid of interest is a urea-based liquid intended to reduce nitrogen oxides in a so-called 'SCR' system mentioned above. The urea-based liquid is injected into a catalytic reducer.
[0064] According to another example, the fluid of interest is a hydrocarbon-based liquid intended to be injected into a combustion chamber.
[0065] As illustrated in [Fig. 1], the drive system involved in the present invention comprises at least one fluid reservoir 3, a pump 2, and a fluid injection device 1 of the type described above. A conduit 21 is provided to fluidly connect the pump outlet to the injection device so as to channel the pressurized fluid PI from the pump to the injection device 1.
[0066] A fluid flow path is thus defined from the pump to an injected zone 4 downstream of the injection device. The pressure downstream of the injection device is denoted PO. System - general information
[0067] The system further comprises an electronic control unit 5 capable of controlling the opening of the injection device. In the case of a needle injector 15, the electronic control unit 5 controls an injector coil 13.
[0068] Pump 2 can be integrated into reservoir 3. Alternatively, pump 2 can be separated from the reservoir.
[0069] A pressure sensor 6 is provided to measure the pressure PI at the pump outlet. This pressure sensor is arranged at the pump outlet or on the pipe 21.
[0070] In the intended SCR-type applications, the PI pressure is between 2 bar and 8 bar. In other applications, the PI pressure may be between 5 bar and 100 bar. Higher pressures are also not excluded from the scope of the present invention, for example, up to 500 bar.
[0071] According to one implementation, the pump 2 is controlled in a continuous mode, namely it runs before the injector is activated and continues to run during the injector activation, and even after the injector is activated. According to another implementation, the pump is controlled in an 'on-demand' mode, namely just before and during the injector activation cycle.
[0072] For example, in SCR applications, it is common practice to use a series of closely spaced injection cycles to generate combustion of nitrogen oxides. Following this, the filter is regenerated and it is no longer necessary to inject urea; consequently, the injector can remain closed at rest for several minutes or for a distance of several kilometers traveled by the vehicle. Therefore, the pump can remain off between groups of injection cycles.
[0073] Another pressure sensor 7 is provided to measure the PO pressure in the injected area.
[0074] Optionally, a temperature sensor 8 is provided to measure the temperature of the fluid, for example the temperature of the fluid in the tank 3.
[0075] In a first configuration, the pipe 21 serves a single injection unit.
[0076] In a second configuration, the pipe 21 forms a manifold and serves a plurality of injection devices, which are controlled in sequential mode, i.e. one at a time.
[0077] The conduit 21 is devoid of a component allowing direct measurement of a liquid flow rate or quantity delivered by the injection component.
[0078] The following section focuses on how best to determine a flow rate value for the fluid of interest in a vehicle of interest. In one example, the vehicle of interest is a specific vehicle among a set of vehicles manufactured in medium or large quantities. This vehicle of interest includes the engine system described above, with a reservoir, pump, and injection component(s).
[0079] In addition, the electronic control unit 5 includes a physical model-based algorithmic computing module (PM), this module is identified as 51 in [Fig. 5]. Loss estimation module learning
[0080] Prior to this, tests and measurements were carried out on a test vehicle similar or comparable to the vehicle of interest, focusing on the flow path. These tests are intended to train a supervised learning loss estimation module 52 which will be used on the vehicle of interest.
[0081] These learning tests are carried out using a series of test injection devices having known and varied flow cross-section characteristics. The test injection devices may have a flow cross-section smaller than the nominal flow cross-section, or conversely, a flow cross-section larger than the nominal flow cross-section.
[0082] One of the test injection units is successively installed as the injection unit on the test vehicle to measure the hydraulic loss introduced by the test injection unit. For each test injection unit, it is activated and a set of parameters is measured. This set of parameters includes at least: - PI pressure - PO pressure - Pressure jump dP upon closing the injection device. This pressure jump is called 'water hammer' in hydraulic jargon. It is a fairly short transient. - Temperature of fluid TKi (TK1, TK2, TK3 taken respectively at different locations). - QMES Measured quantity of fluid actually injected.
[0083] The measured quantity QMES can be effectively measured because these tests and trials are carried out with laboratory means on the test vehicle.
[0084] The test injection element simulates non-nominal behavior on the flow path in the vehicle of interest as a function of the plurality of parameters.
[0085] Each of the test injection devices has a different passage section so as to cover fairly broadly the spectrum of passage section evolutions that can be encountered on production vehicles.
[0086] In one example, a supervised learning loss estimation module is used. A test injection device is installed, and the injected quantity and the parameters dP, Pl, P0, and T are measured. This process is then repeated with another test injection device. For an output of the estimation module called the hydraulic loss coefficient, denoted CP, a cost function is calculated and minimized as the learning progresses.
[0087] Thus, in operational mode, the loss estimation module will take as input said plurality of parameters and will provide as output the hydraulic loss coefficient noted CP.
[0088] The loss estimation module includes a neural network. The coefficients of the neural network nodes are adjusted by a backpropagation learning process based on a cost function. For example, the cost function can be based on QMES - CP x QTH
[0089] As illustrated in [Fig. 1], the size of the neural network is moderate. In terms of structure, the neural network can comprise two hidden layers, with 10 to 12 neurons for each hidden layer 42,43. Generally, the total number of neurons can be chosen to be less than 500, preferably less than 400. Thus, the computation time is very small and the computation repetition frequency can be high.
[0090] Also, the memory size occupied by the neural network is very modest. The neural network can be less than 5 kilobytes in size and prefers- significantly less than 3 kilobytes.
[0091] However, the principle of the invention can be used for larger neural network sizes. Operational appraisal of a vehicle of interest
[0092] The electronic control unit is configured to implement the following steps:
[0093] / b / to carry out a plurality of N fluid injection sequences, during which one collection of values of a plurality of parameters (dP,Pl,P0,T,X),
[0094] / c / - calculate a theoretical quantity of fluid injected during these N sequences fluid injection, using at least some of the said values of the plurality of parameters (Pl,P0,T,X),
[0095] / d / - transmit, to the loss estimation module, said values of the plurality of parameters (dP,Pl,P0,T,X), and obtain, at the output of the loss estimation module, the loss coefficient CP,
[0096] / e / - calculate an estimated actual quantity of fluid injected during the N sequences fluid injection, by applying the loss coefficient CP to the calculation of the theoretical quantity of fluid.
[0097] In step / a / , the values of the different parameters are sampled by the pressure sensors 6, 7 and by the temperature sensor 8. Sampling can be quite rapid, particularly for the pressure measurement PI. Conversely, temperature measurements do not require rapid sampling.
[0098] For measuring the PI pressure, in order to capture the characteristics of the water hammer, the sensor must be a high-speed sensor, and the sampling rate must allow for at least 50 samples per second (at least 50 kHz). In one example, the sampling rate allows for at least 100 samples per second (at least 100 kHz on the capture and digitization chain). Figure 8 illustrates an example of an overpressure waveform corresponding to the water hammer at injector closure. In reality, the characteristics of this waveform reflect the state of the liquid passage cross-section. Optionally, other additional parameters Y can be taken into account, such as the time elapsed since the previous injection series, the outside temperature, the vehicle mileage, etc.
[0099] According to an example, the theoretical quantity of fluid, denoted QTH, can be calculated as follows, using a so-called Bemouilli model for an incompressible fluid:
[0100] [Math.2] QTH = Sil iTi A
[0101] In this formula:
[0102] p is the density or mass density of the fluid,
[0103] PI is the pump outlet pressure,
[0104] PO is the pressure in the injected zone 4,
[0105] A is a characteristic section of the passage section,
[0106] Ti is the injection duration, for one control cycle of the injector, index i,
[0107] N the number of injections.
[0108] QTH is a mass quantity. This quantity corresponds to the nominal quantity for a new injector. This calculation corresponds to step / c / of the promoted process. QTH is calculated by the MP 51 module.
[0109] It should be noted that the characteristics of the water hammer do not appear in the above formula; the pressure PI is an average pressure and the pressure jump only plays a marginal role in the above formula.
[0110] Conversely, in step / d / , the loss estimation module takes as inputs not only the PI,PO parameters mentioned above but also the characteristics of the pressure jump dP (water hammer). The characteristics of the pressure jump dP include at least the jump height dPl and the duration Tr, as illustrated in [Fig.8].
[0111] The loss estimation module, whose supervised learning was previously carried out, now delivers its output in the form of the loss coefficient CP.
[0112] According to a simple example, the estimated actual quantity of injected fluid, denoted QRE, is calculated by multiplying the theoretical quantity of fluid, denoted QTH, by the loss coefficient CP from the loss estimation module. Therefore, in this case, QRE = CP x QTH.
[0113] According to a generic formulation, the estimated actual quantity of fluid is calculated by applying the loss coefficient CP to the calculation of the theoretical quantity of fluid according to a correction function F, as, for example, expressed as follows:
[0114] QRE = F(QTH, CP).
[0115] If N injections are performed, each having an index 'i' and a duration Ti, the theoretical and corrected actual quantities (QTH,QRE) are calculated as a summation over the index 'i'.
[0116] If all the Ti values are the same, then the calculation of QTH simplifies to:
[0117] [Math.3] QTH = N Tî dp(Pl - PQ) A,
[0118] In practice, PI is taken to be either the mean or the median of the values collected over the N injections. Similarly, PO is taken to be either the mean or the median of the values collected over the N injections. Likewise, the characteristics of the pressure jump dP, i.e., dPl and Tr, are taken to be either the mean or the median of the values collected over the course of the N injections.
[0119] Depending on an option, an alert is activated if the loss coefficient CP is less than a predetermined threshold CPS1. For example, the CP1 threshold is preferably between 0.5 and 0.75. Control system
[0120] The electronic control unit 5 includes a microcontroller, a non-volatile memory area, analog-to-digital converters for acquiring pressure and temperature parameters.
[0121] As seen in [Fig.1], the electronic control unit 5 includes at least one output 55 for controlling the injector coil 13.
[0122] As shown schematically in [Fig.5], the electronic control unit 5 performs the calculation of the theoretical quantity injected, based on the opening times Ti. This is represented as the module 51 already discussed and visible in [Fig.5].
[0123] Furthermore, the electronic control unit 5 includes the loss estimation module 52. For each unit injection cycle or each group of injection cycles, the loss estimation module provides the loss coefficient CP and the electronic control unit 5 uses this output to generate an alert for the driver or the vehicle maintenance service if necessary.
[0124] The electronic control unit 5 performs the calculation of the estimated actual quantity (step / e / ). All the steps are illustrated in [Fig.6].
[0125] Regarding the neural network 40, illustrated in [Fig. 7], it preferably has a simple structure, with a one-dimensional output, in this case the coefficient CP. The input layer comprises a number of parameters XI, X2, X3, Xi, Xm. The number m of inputs can be on the order of 10.
[0126] The neural network 40, includes 1 to 3 intermediate layers 42 43, also called hidden layers, for example of the same dimension as the input vector.
[0127] Thus, the size of the neural network 40 is reduced. For example, the number of nodes can be around 30 and the number of neurons / links close to 300, and therefore less than 400 as already mentioned above. Each parameter can be stored in 4 bytes, resulting in the modest memory sizes mentioned above. Other points
[0128] The test vehicle may be of the same type as the target vehicles where the process described above is used on a large scale. However, it should be noted that the test vehicle may be of a different type than the target vehicles, insofar as the flow path and the injection device exhibit a certain similarity between the test vehicle and the target vehicles.
Claims
Claims
1. Method for determining an injected quantity of a fluid of interest in a motorization system of a vehicle of interest, said quantity not being measured directly by a sensor, the motorization system comprising at least one fluid reservoir (3), a pump (2), a fluid injection member (1), with a fluid flow path from the pump to an injected zone (4) downstream of the injection member, and an electronic control unit (5) capable of controlling an opening of the injection member otherwise closed in the absence of a command, the method comprising the following steps: - having a loss estimation module (52) with supervised learning (RNN, IA), taking as input a plurality of parameters (dP, Pl, P0, T, X) and providing as output a hydraulic loss coefficient CP relating to a hydraulic loss introduced by the fluid injection member (1), / b / - carrying out a plurality of N fluid injection sequences, during which values of said plurality of parameters (dP, Pl, P0, T, X) are collected, / c / - calculating a theoretical quantity (QTH) of fluid injected during these N fluid injection sequences, using at least part of said values of the plurality of parameters (PI, PO, T, X), the theoretical calculation using a so-called Bernoulli model for an incompressible fluid, / d / - transmitting, to the loss estimation module (52), said values of the plurality of parameters (dP, Pl, P0, T, X), and obtaining, at the output of the module loss estimation, the loss coefficient CP,ld- calculate an estimated actual quantity (EAQ) of fluid injected during the N fluid injection sequences, by applying the loss coefficient CP to the calculation of the theoretical quantity of fluid.,
2. Method according to claim 1, comprising a preliminary step: / a / - carrying out a preliminary training of the supervised learning loss estimation module (RNN, IA) by means of a series of test injection members having known passage section characteristics, which are placed in turn as injection members on a similar flow path of a test vehicle to simulate the hydraulic loss introduced by the injection member on the flow path in the vehicle of interest as a function of the plurality of parameters, and by measuring during an opening sequence of the injection member, the values of the parameters of the plurality of parameters, the loss estimation module taking as input said plurality of parameters (dP, Pl, P0, T, X) and providing as output the hydraulic loss coefficient CP.
3. Method according to claim 1 or 2, in which in step / e / the loss coefficient CP is applied by multiplying it to the calculation of the theoretical quantity (QTH) of fluid to obtain the estimated actual quantity (QRE) of fluid injected during the N fluid injection sequences.
4. Method according to one of claims 1 to 3, in which the loss coefficient CP is between 0 and 1.
5. Method according to one of claims 1 to 4, in which an alert is activated if the loss coefficient CP is lower than a predetermined threshold, which is preferably between 0.5 and 0.
75.
6. Method according to one of claims 1 to 5, in which the loss estimation module comprises a neural network, and preferably the neural network occupying a memory size of less than 5 kilobytes.
7. Method according to one of claims 1 to 6, in which the values of the plurality of parameters (dP,Pl,P0,T,X) are filtered and / or smoothed on the N injection sequences for use in the loss estimation module.
8. Method according to one of claims 1 to 7, in which the plurality of parameters (dP,Pl,P0,T,X) comprises a first parameter (dP) representative of an increase in pressure when the injector is closed.
9. System for determining an injected quantity of a fluid of interest in a motorization system of a vehicle of interest, said quantity not being measured directly by a sensor, the motorization system comprising at least one fluid reservoir (3), a pump (2), a fluid injection member (1), with a fluid flow path from the pump to an injected zone (4) downstream of the injection member, and an electronic control unit (5) capable of controlling an opening of the injection member otherwise closed in the absence of a command, the fluid of interest being an incompressible liquid fluid or of low compressibility, the system comprising a hydraulic loss estimation module (52) with supervised learning (RNN, IA), taking as input a plurality of parameters (dP, Pl, P0, T, X) and providing as output a hydraulic loss coefficient CP relating to a hydraulic loss introduced by the fluid injection member (1), the electronic control unit (5) being configured to: / b / - carry out a plurality of N fluid injection sequences, during which values of said plurality of parameters (dP, Pl, P0, T, X) are collected, / c / - calculate a theoretical quantity (QTH) of fluid injected during these N fluid injection sequences, using at least part of said values of the plurality of parameters (PI, PO, T, X), the theoretical calculation using a so-called Bernoulli model for an incompressible fluid, / d / - transmit, to the loss estimation module, said values of the plurality of parameters (dP, Pl, P0, T, X), and obtain, at the output of the module loss estimation, the loss coefficient CP,ld- calculate an estimated actual quantity of fluid injected during the N fluid injection sequences, by applying the loss coefficient CP to the calculation of the theoretical quantity of fluid.,
10. System according to claim 9, in which provision is made for training the supervised learning loss estimation module (RNN, IA), prior to actual use, by means of a series of test injection members having known passage section characteristics, which are placed in turn as injection members on a similar flow path of a test vehicle to simulate the hydraulic loss introduced by the injection member on the flow path in the vehicle of interest as a function of the plurality of parameters, and by measuring during an opening sequence of the injection member, the values of the parameters of the plurality of parameters, the loss estimation module taking as input said plurality of parameters (dP, Pl, P0, T, X) and providing as output the hydraulic loss coefficient CP.
11. A system according to any one of claims 9 to 10, wherein the fluid is a urea-based liquid for reducing nitrogen oxides.
12. System according to one of claims 9 to 11, in which the injection member is a needle injector.
13. System according to one of claims 9 to 12, in which the loss estimation module is implemented in the form of a neural network. included in the electronic control unit (5).
