FR3000206A1 - Method for characterizing operation of internal combustion engine i.e. compression ignition engine from sequence of engine tests, involves deducing value representative of stabilized signal from sigmoidal function, and analyzing value - Google Patents

Abstract

The method involves acquiring, for each engine test, a signal from a continuous measurement of an operating parameter of an internal combustion engine over an acquisition period, where the acquisition period is selected to determine a sigmoidal function to model the signal. The sigmoidal function for modeling the signal is determined, and a value representative of a stabilized signal is deduced from the sigmoidal function. The value representative of the stabilized signal is analyzed to characterize the operation of the internal combustion engine. Independent claims are also included for the following: (1) a computer program product for characterizing the operation of an internal combustion engine (2) a test device for performing a sequence of engine tests for an internal combustion engine.

Description

Field of the invention The present invention relates to tests on internal combustion engines to characterize their operation, for example for calibration purposes, said tests being for example carried out on a motor bench of any type - stationary, dynamic or high - dynamic - or performed on the vehicle in operation. General context Motor tests consist in measuring and recording the responses of an engine placed in particular operating conditions, these conditions being simulated or real, and for example corresponding to specific driving conditions of a vehicle in environmental terms ( outside air temperature, altitude, slope of the road) or in terms of use (urban or highway rolling, sporting or economic behavior, mass towed by the vehicle). The information obtained then makes it possible to characterize the operation of the engine in relation to the particular conditions applied, and can for example be used to calibrate the variables of the engine control strategies programmed in the electronic calculator in particular to meet various standards and criteria. , such as standards for pollutant emissions, noise, flexibility of driving, etc.

These tests are traditionally performed on engine test benches, also called engine test benches or engine benches, for example during engine design phases, adjustment before putting on the market, or throughout the industrial production of the engine. engine. Each test generally consists of numerous measurements of different operating parameters of the engine or its members, these parameters reflecting their operating state, in particular the state of the combustion of the engine and the thermal, mechanical, kinetic and chemical consequences of this. reaction. The operating parameters are, for example, concentrations of chemical compounds in the exhaust gases, for example polluting substances, the pressure and the temperature of these gases, the noise generated by the combustion in the engine, the rotation speed of the engine. turbo-compressor shaft, etc. During an engine test sequence, the engine settings (injection pressure, injection timing, air pressure, etc.) and / or the operating point of the engine (FIG. diet, load). The operating point of the motor corresponds to a torque (speed, load), and will also be referred to as the power point in the present description, in relation to the power of the engine, which is expressed as the product of the rotational speed of the engine. (regime) by the couple (load). Among the measurements made during the motor test, it is possible to distinguish the continuous measurements from the point measurements, the continuous measurements representing the vast majority of the measurements generally made. Continuous measurements consist of measuring the response of the motor following each variation of the settings and / or the power point of the motor. Continuous measurements are measurements made by so-called synchronous devices, performing measurements at regular intervals and having a high availability rate. This type of measurement is performed with a relatively fast sampling rate (typically 1 to 10 Hz): the observations are refreshed in time every second or even more often. Continuous measurements are, for example, permanent measurements, which are either averaged over a sliding time window, or raw, such as pressure, temperature, velocity, voltage, flue opacity, cylinder pressure measurements. allowing to establish quantities describing the combustion. Other continuous measures, such as the measurement of chemical species concentration of flue gases, the instantaneous measurement of fuel consumption, can sometimes be interrupted during purge operations, allowing the removal of certain chemical compounds impregnating the fuel lines. gas sampling to restore the quality of measurement and avoid drifts, or during filling operations (some devices consist of a bucket placed on a scale to weigh the fuel that flows into the engine). Point measurements are performed by so-called asynchronous devices that perform an observation at a given time and over a limited time. After each observation, the value does not change over time. This may be the measurement of the smoke index by exposure of a filter to a volume of exhaust gas taken from the exhaust, the integrated measurement of fuel consumption, carried out en masse over a period of time or in time for a given mass of fuel consumed.

The set of measurements during an engine test sequence then makes it possible to characterize the operation of the engine under the particular conditions of the tests. Traditionally, a continuous measurement during an engine test, that is to say the continuous measurement of a quantity for setting parameters and / or a given power point of the engine, is broken down into three stages, such as FIG. 1 shows the measurement. The measurement begins with a first phase 11 of rapid modification (in a few seconds at most, represented by the time interval t1-t2 in FIG. 1) of the engine settings (PR), for example the fresh air flow at the intake (Qaid or the amount of fuel injected for the pilot injection (Qin; Pi!), or its power point, for example the speed and / or the load. In response, the acquired signals of the operating parameters PF of the motor generally vary strongly The next phase is a waiting phase of the stabilization of the signal 12 (time interval t243), which must be the same for all the measurements made because generally objective is to measure phenomena that are At the same time, it ends when all motor operating parameters have a smaller variation than a parameter-specific threshold (a few percent) for a given time (about 10 seconds). In a test bench test, this second phase typically lasts 45 seconds to obtain a good quality of measurement.

A third phase corresponds to a signal averaging phase 13. This third measurement phase 13 makes it possible to calculate a value representative of the continuously stabilized continuous measurement, by averaging all the samples recorded in the corresponding time interval t3. t4. It is common, in a bench test, that this phase lasts 30 seconds to obtain a reliable result.

The continuous measurement carried out according to the different phases described during an engine test is known in the field of motor tests under the name of point taking. An engine test sequence thus makes it possible to record synthetic information, for example in a computer file, where for each continuous measurement made during the engine test corresponds a single value representing the average of all the samples of this measurement. continuous over the time interval of the averaging phase 13. The values are indexed according to a point number. This synthetic information can then contribute to establishing a database relating to the engine test sequence carried out, this database being able to be processed with digital tools so as to provide additional information concerning the operating state of the engine, such as that statistical behavioral models (governed for example by simple functions such as polynomials) or physical (governed by systems of complex equations to solve). A major problem with this traditional stitching is that the stabilization and averaging phases are very time-consuming during the engine test. As shown in Figure 1, these two phases account for more than half of the duration of a continuous measurement sequence during the motor test, and often more than 80% of the total duration of the motor test . It is very difficult to reduce the duration of the stabilization phase without taking the risk of exposing the next phase of averaging to thermomechanical phenomena having different inertia and / or dynamics that could lead to invalidating this averaging phase, and ultimately result in having to repeat the measurement and, therefore, lengthen the duration of the test. Similarly, reducing the duration of the averaging phase can introduce a bias in the result.

Another problem encountered during this point acquisition is that the averaging phase, uniquely determined for all continuous measurements, does not always provide a representative value of the stabilized signal. Indeed, there may be some measures that still vary, where others have stabilized. This is for example the case of the measurement of the concentration of unburned hydrocarbons (HC) in Figure 1, which continues to evolve during the averaging phase while the measurement of the concentration of carbon monoxide (CO) is stabilized for much longer. This kind of phenomenon is relatively common during motor tests.

The present invention generally aims to reduce the duration of motor tests, and more particularly the waiting time associated with signal stabilization and averaging phases, traditionally required to calculate a representative value of a signal once stabilized. Another object of the invention is to provide a reliable method for characterizing the operation of an internal combustion engine during an engine test sequence, in that a value representative of the stabilized signal is provided for each continuous measurement. For this purpose, the invention proposes a method for characterizing the operation of an internal combustion engine, based on an engine test sequence, each engine test comprising the modification of at least one engine control parameter and or its power point, comprising the following steps: a) for each engine test: i. at least one signal is acquired from the continuous measurement of an operating parameter PR of said engine over an acquisition period d, the acquisition duration d being chosen so as to be able to determine at least one sigmoid function for modeling the said at least one signal; ii. said at least one sigmoid function modeling said at least one signal; iii. a representative value of said at least one stabilized signal is deduced from said at least one sigmoid function; b) analyzing all of said representative values of the stabilized signals to characterize the operation of said internal combustion engine. Advantageously, at least one adjustment parameter PR and / or the power point of the motor is varied rapidly, preferably between 0 and 5 seconds, after the triggering of the signal acquisition.

Preferably, the continuous measurement is carried out at a sampling frequency of between 1 Hz and 10 Hz. Preferably, said at least one sigmoid function is a Gompertz function.

The graphical representation of said at least one sigmoid function preferably comprises a first horizontal asymptote whose value is determined at an initial instant te, of the signal and a second horizontal asymptote whose value is determined at an infinite instant t. embodiment, the signal is modeled by a sigmoid function written according to the following equation (I): y (t) = yo + ysiep x exp (-bx exp (-cx (t -0)) (I) comprising the five parameters to be determined yo, Ystep, b, c and tb YO being the value of the operating parameter PF at an initial moment t = to; ystep being the difference between the value of y (t) at an infinite time t = t and to, b and c being positive constants, with b characterizing the shift of the curve in time and c characterizing the slope of the curve, ti being a positive time constant, according to one embodiment, the signal by two sigmoid functions.

Preferably, the two sigmoid functions are combined under the following equation (II): y (t) = yo + y 'epi x exp (-b1x exp (-c1x (t-t11))) ± Ystep2x exP (- b2 x exp (-c2x (t-ti2))) (11) comprising the nine parameters to be determined Yo, v stepl, Ystep2, b1, c1, ti1, b2, C2e ti2 yo being the value of the operating parameter PF at one initial time t = t0; Ystepl being the difference between the value of y (t) at infinite time t = t- and that at to if a single first thermomechanical and / or thermo-chemical phenomenon existed, and representing the variation of the signal related to this a first phenomenon; Ystep2 being the difference between the value of y (t) at infinite time t = t if only the first phenomenon existed and the value of y (t) at infinite time t = t- if only one second thermomechanical phenomenon and / or thermo-chemical antagonist to the first phenomenon existed, Ystep2 representing the variation of the signal related to this second phenomenon; Ystepl and Ystep2 being of opposite signs; y, and c1 being positive constants, with b1 characterizing the shift of the first S-curve in time, and c1 characterizing the slope of the first S-curve; being a positive time constant relative to the inertia of the first phenomenon; b2 and c2 being positive constants, with b2 characterizing the shift of the second S-curve in time, and c2 characterizing the slope of the second S-curve; ti2 being a positive time constant relative to the inertia of the second phenomenon. Step ii) of determining said at least one sigmoid function modeling the signal can comprise the following steps: the signal having a dynamic criterion or a time evolution criterion greater than or equal to a predetermined dynamic threshold value is selected or a predetermined evolution threshold value (QUAL-S); the parameter yo corresponding to the value of the first horizontal asymptote (Val-AS1) is determined; the time derivative of the signal (CH-FS) is calculated; the sigmoid function is defined according to equation (I) or the two sigmoid functions according to equation (II) according to whether the sign of the temporal derivative of the signal remains the same or changes (CH-FS); the other parameters of said sigmoid function (s) are calculated in such a way as to minimize, in the least squares sense, the difference between the acquired signal and the estimated signal according to the said sigmoid function (CAL-PFs); and step iii) of determining the representative value of the stabilized signal may comprise the following steps: the value of the second asymptote y is determined. (Val-AS2); a correlation coefficient is calculated between the acquired signal and the sigmoid function (s) and y is selected as a representative value of the stabilized signal if the correlation coefficient is greater than a predetermined correlation value (QUAL-corr, CAL- FP).

Advantageously, the acquisition time d is increased at the end of the step iii) of determining the value representative of the stabilized signal so that the correlation coefficient is greater than a predetermined correlation value. The method may further comprise a signal filtering step (FIL) during step ii) of determining the sigmoid function modeling the signal, said filtering comprising a preference filter chosen from the following list of filters: moving average, Butterworth filter, first-order low-pass filter. Advantageously, the engine test is performed on a test engine bench or on a vehicle in operation. Advantageously, the internal combustion engine is calibrated from the analysis of all the representative values of the stabilized signals characterizing the operation of said engine.

The method is preferably used to characterize the operation of a compression ignition engine. The invention also proposes a computer program product for keeping in a memory of a unit of a processor or on a removable memory medium suitable for cooperating with said processor unit, the program product comprising instructions for implementing the process according to the invention. The invention also proposes a device for testing an internal combustion engine that makes it possible to carry out an engine test sequence, comprising: measuring instrumentation for the continuous measurement of at least one PF operating parameter of said engine generating a raw time signal during an engine test; an automation system for supervising the carrying out of the motor test and for displaying and / or recording the signals acquired with the measurement instrumentation; a computer program according to the invention for determining a value representative of said stabilized signal; Said automation system being able to display and / or record all the representative values determined by the computer program, for their analysis to characterize the operation of said combustion engine; and said computer program being able to be integrated into said automation system or to said measurement instrumentation, or else to a computer separate from said automation system and from said measurement instrumentation, said computer being in communication with said automation system . Other features and advantages of the invention will appear on reading the following description of nonlimiting examples of embodiments, with reference to the following appended figures. FIG. 1 is a diagram schematically illustrating a traditional continuous measurement sequence (point taking) of operating parameters PF of an internal combustion engine. FIG. 2 is a diagram showing the temporal evolution of the CO2 emitted in the exhaust gases of an internal combustion engine during an engine test sequence. Each point of the diagram corresponds to a value representative of the concentration of CO2 during a motor test (Pts) calculated according to the prior art of the invention (points symbolized by gray circles) and according to the invention (points symbolized by black squares). FIG. 3 corresponds to the raw time signal (amount of CO2 measured over 220 seconds) making it possible to estimate the representative value (a) of FIG. 2, and shows the signal modeled according to the invention taking into account the first 60 seconds. of the raw signal. FIGS. 4A, 4B and 4C represent a raw time signal corresponding to the continuous measurement of the concentration of unburned hydrocarbons (HC) during the same engine test sequence as that of FIGS. 2 and 3, and represent the modeling of this signal by a double sigmoid curve according to the invention from the first 60 seconds of signal. FIG. 5 is a flowchart describing the calculation of a value representative of a stabilized signal according to the invention.

FIGS. 6A and 6B are examples of non-qualified time signals for modeling with a sigmoid function according to the invention. FIGS. 7A and 7B illustrate two embodiments of the invention: FIG. 7A illustrates a modeling of the signal according to a sigmoid function and FIG. 7B illustrates the modeling of the signal by two sigmoid functions.

FIGS. 8A, 8B and 8C are diagrams comparing the evolution of the amount of CO2 as a function of time calculated according to the prior art and according to the invention (FIG. 8A) and representing the performance of the process according to the invention (FIGS. 8B and 8C). FIGS. 9A, 9B and 9C are diagrams comparing the evolution of the concentration of unburned hydrocarbons (HC) as a function of time, calculated according to the prior art and according to the invention (FIG. 9A), and representing the performance of the process according to the invention (Figures 9B and 9C). Figure 10 is a diagram illustrating the implementation of the method according to the invention in a motor bench.

By engine operating parameter (s), denoted by PF in the present description, is meant parameters that can correspond to a physical or chemical quantity making it possible to obtain information on the operation of the engine and thus reflecting an operating state of the engine or of his organs. It may be a concentration of a compound in the exhaust gas, such as carbon monoxide (CO), carbon dioxide (CO2), unburnt hydrocarbons (HC), different nitrogen oxides (NOx), the temperature or pressure of the gases at the inlet or the exhaust, the maximum pressure of the gases in the combustion chamber, an angle of the crankshaft at which a fraction of the mass of fuel burned is located, a calculated combustion noise of after a cylinder pressure signal, a fuel flow, a rotational speed of a turbo-compressor shaft, the richness of the gases at the intake or the exhaust, etc. Engine adjustment parameter (s), denoted by PR in the present description, are understood to mean parameters such as an air flow, a pressure of the gases at the intake, a quantity of fuel injected for each injection. , a fuel injection pressure, a separation time between different injections, a level of swirl in the combustion chamber, or any other parameter to change the course of combustion and consequently the behavior of the engine.

According to the invention, it is possible to characterize the operation of an internal combustion engine, in particular a compression ignition engine, from a motor test sequence, each test comprising the modifications of at least one setting parameter of the PR motor and / or its power point.

For this, the applicant proposes a method comprising the following steps: a) for each engine test: I. at least one signal is acquired from the continuous measurement of an operating parameter PF of the engine over an acquisition period d, the acquisition duration d being chosen so as to be able to determine at least one sigmoid function for modeling said at least one signal; ii. said at least one sigmoid function modeling said at least one signal; iii. a representative value of said at least one stabilized signal is deduced from said at least one sigmoid function; b) analyzing all the representative values of the stabilized signals to characterize the operation of said internal combustion engine. The set of representative values of the stabilized signals resulting from an engine test sequence comprising continuous measurements of one or more PF operating parameters of the engine makes it possible to establish the temporal evolution of the parameter or parameters during the test. . It is the analysis of this evolution of the parameter (s) throughout the duration of the engine test sequence (sum of engine test durations) which makes it possible to characterize the operation of said combustion engine.

Figure 2 illustrates such a time evolution for the CO2 concentration in the exhaust gas during a motor test sequence. The evolution of the CO2 concentration is represented by the value of the CO2 concentration (in percentage) as a function of the motor tests indexed on the abscissa, also called "points" (points 9330 to 9520). Each measurement point, symbolized by a square, corresponds to the representative value of a stabilized signal calculated according to the invention, that is to say to the estimated stable value of the signal modeled by at least one sigmoid function according to FIG. invention. The gray circles symbolize the values calculated according to a traditional point catch, where the representative value corresponds to an average performed on a phase of averaging the signal. Thus, for a given number n of signals acquired relative to an operating parameter of the engine PR during a sequence of n engine tests, n being sufficient to characterize the operation of the engine, n representative values (or measurement points) are calculated. allowing then to determine the evolution of the operating parameter PR as a function of time, each acquired signal, also commonly called "point", having an acquisition time d. In Figure 2, the evolution of CO2 over time is achieved through 191 signals acquired during 191 engine tests (index 9520 - index 9330), represented in 10 abscissa under the name "points", giving 191 points of measurement. The engine test sequence shown in Figure 2 includes the following features: The test sequence combines 191 rapid and simultaneous power point variations and experimental plane settings. The total duration of the test sequence is 9.45 hours, or about 184 seconds per unit engine test. The engine tested is a 15 diesel engine (PSA DV6 Turbo-Diesel Injection Direct Common-Rail). The tests are carried out on a dynamic engine bench. The acquisition of all the signals is carried out at a frequency of 1Hz. The following operating parameters are measured: - pollutants: CO2, CO, HC, NOx, 02, recirculated CO2 at intake; - voltage delivered by a wealth probe; 20 - instantaneous calculations: richness "5gaz" (richness calculated from the concentrations of five gases), combustion noise (Noisemeter system), angle of the fraction of gases burned at 50%, maximum cylinder pressure; kinetics: speed of rotation of the turbo-compressor; temperatures: intake gas, exhaust gas; 25 pressures: intake gas, exhaust gas; flow rates: fresh air at the intake, fuel (via a KMA4000 system). The graphical representation of the sigmoid function or functions modeling the signal preferably comprises a first horizontal asymptote whose value is determined at an initial moment of the signal 1.0 and a second horizontal asymptote whose value is representative of a moment which would be located at infinity to 0. The initial time to corresponds to the first samples of the time signal, when the PR adjustment parameters and / or the motor power point have not yet been modified. The infinite moment L corresponds to the samples of the stabilized signal. Thus, the second asymptote corresponds to the stabilized part of the signal, that is to say the part of the signal where the variations are no longer significant, that is, they are smaller than a threshold specific to each parameter, for example a few percent, for a given duration, for example for about 10 seconds. According to the invention, the time signal is modeled by at least one sigmoid function. Preferably, the signal is modeled by at least one sigmoid function which is a Gompertz function. According to one embodiment, the signal is modeled by a sigmoid function written according to the following equation (I), which is a function of Gompertz: y (t) = yc, + Ystep x exp (-bx exp (- cx (t - t1))) (I) with: yo being the value of the operating parameter PF at an initial time t = to; Ystep being the difference between the value of y (t) at infinite time t = t ", and that at an initial time to; b and c being positive constants, with b characterizing the shift of the curve over time, and c characterizing the slope of the curve, and; t being a positive time constant, and indicating that the inertia of the physical phenomenon (impact of the modification of an engine control parameter and / or its power point on an engine operating parameter) is not necessarily linked to to. In this case, the function comprises five parameters to be determined yo, Ystep, b, c and ti; FIG. 3 illustrates the graphical representation of a sigmoid function according to equation (I), modeling the signal resulting from the measurement of the CO2 concentration in the exhaust gases during a motor test (indexed as Pt 9382) of the engine test sequence of FIG. 2. FIG. 3 shows the raw time signal acquired over 220 seconds, and the S curve modeling such a signal determined according to the invention, making it possible to provide the representative value (a). symbolized by a square in Figure 2. A duration of 60 seconds is sufficient to establish the sigmoid function modeling the acquired signal. According to another embodiment, the signal is modeled by two sigmoid functions, the graphical representation of which may comprise two opposite sigmoidal curves of variation. Such a double S-shaped curve with opposite directions of variation is illustrated in FIG. 3 and FIGS. 4A, 4B and 4C. The curve FS of FIG. 4A is superimposed on a signal representing the continuous measurement over time of the HC concentration during an engine test (motor test indicated by point 9501) of the test sequence described above in FIG. relation with Figures 2 and 3. Preferably, the two sigmoid functions are Gompertz functions and express themselves under the following equation (Il): y (t) = - yo + y step x exp (-b1x exp ( -c1x (t - ti1))) + Ystep2 x exp (-b2x exp (-c2x (t-t12))) (Il) with: yo being the value of the operating parameter PF at an initial moment t = to; ystepi being the difference between the value of y (t) at an infinite time t = t., and that at an initial time to if a single first thermomechanical and / or thermo-chemical phenomenon existed and representing the variation of the signal related to this first phenomenon; Ystep2 being the difference between the value of y (t) at infinite time t = t., If only the first phenomenon existed and the value of y (t) at infinite time t = you if a single second thermo phenomenon mechanical and / or thermo-chemical antagonist to the first phenomenon existed, Ystep2 representing the variation of the signal related to this second phenomenon; Ystepl and Ystep2 being of opposite signs; 1) 1 and c, being positive constants, with bl characterizing the shift of the first S-curve in time, and c1 characterizing the slope of the first S-curve; t11 being a positive time constant relative to the inertia of the first phenomenon; b2 and c2 being positive constants, with b2 characterizing the shift of the second S-curve in time, and c2 characterizing the slope of the second S-curve; t12 being a positive time constant relative to the inertia of the second phenomenon; This second function comprises nine parameters to be determined v, v stepl, Ystep2, 131, C1, b2, c, ti2; The modeling of the signal by such a double sigmoid function with a double S curve, the two curves having opposite directions of variation, is particularly well suited to the case where there is a competition between several thermomechanical and / or thermo-chemical phenomena with a inertia and / or a different dynamic and antagonistic. This may for example be the case of continuous measurements of the HC concentration, as shown in the diagrams of Figures 4A, 4B and 4C. The parameters to be determined from these functions make it possible to obtain a curve or several S-shaped curves that fit the shape of the time signals recorded during tests on an internal combustion engine.

Other sigmoid functions able to model the signal can be used in the method according to the invention, such that the sigmoid function expressing itself in the following manner: y (t) = yo + ystep x exp (-b xi), where yo is the value of the operating parameter PF at an initial moment t = to, ystep is the difference between the value of y (t) at infinite time t = t., and that at an initial time to, and b is a positive constant characterizing the shift of the curve over time. FIG. 5 diagrammatically shows what is done for a given signal relating to an operating parameter PF at each engine test, in order to determine a value representative of the stabilized signal relating to a given operating parameter, acquired over a period of acquisition of . As illustrated in Figure 5, eight main steps are performed: - Step 1: Acquisition of a raw time signal from a continuous measurement (ACQS). - Step 2: Qualification of the raw time signal (QUAL-S). - Step 3: Calculation of the value yo (value of the first asymptote) (Val-AS1). - Step 4: Optional step of filtering the raw time signal (FIL). - Step 5: Choice of the sigmoid function (s) and the number of associated parameters (CH-FS). - Step 6: Determine the parameters of the sigmoid function (s) (CAL-PFs). - Step 7: Calculation of the value y. (value of the second asymptote) (Val-AS2). - Step 8: Qualification of the result by calculating a correlation coefficient between the signal and the sigmoid function (s) (QUAL-corr), making it possible to determine (CAL-PF) the representative value of the stabilized signal (measuring point). Steps 2 to 6 make it possible to determine the sigmoid function (s) modeling the signal (raw or filtered), and steps 7 and 8 make it possible to determine a value representative of said stabilized signal from said modeling by the sigmoid function (s) carried out in steps 2 to 6.30 Step 1: Acquisition of a raw time signal from a continuous measurement During the motor test, it is preferably rapidly, for example between 0 and 5 seconds, at least one setting parameter PR of the motor and / or its power point. The power point of the engine can be characterized in particular by the engine speed or the engine load. The engine load can be characterized by different quantities, such as the amount of air introduced and fuel injected, the actual average torque (or indicated if the friction is taken into account) or the effective average pressure (or indicated). This latter quantity corresponds to the constant force exerted on each piston during a complete cycle and which can be connected to the torque by the displacement of the engine. These variations are comparable to a step when they are instantaneous and at a ramp when they last a few seconds. This has the effect of changing the initial conditions of the mixture introduced into the combustion chamber and the course of this reaction and consequently all or part of the operating parameters of the PF motor. When an operating parameter PF evolves significantly, it is observed that the temporal recording of the signal resulting from its measurement has the following characteristic behavior: a plateau precedes the variation which starts slowly then accelerates and slows down before ending by another landing; between acceleration and slowdown, there is a point of inflection. This behavior is most often at work even when the measurement noise masks its presence. The sudden variation of the setting parameter PR of the motor and / or its power point is such that the acquired signal is identifiable by at least one sigmoid function. A time acquisition step of raw signals is performed before any step using such signals as described below. This step makes it possible to record the effects of the variation of at least one adjustment parameter PR of the motor and / or its power point on the operating parameters PF of the engine. This acquisition step consists in recovering at least one signal derived from the continuous measurement of an operating parameter PF of the internal combustion engine, this measurement being carried out using suitable measuring means, and recording this signal. in an acquisition unit. Continuous measurement during an engine test is understood to mean any measurement carried out at regular and short time intervals, preferably carried out at a sampling frequency f between 1 Hertz and 10 Hertz. This frequency range is adapted to the duration of variation of the operating parameters PF which is preferably set to a few tens of seconds. The higher the frequency, the more the signal offers a reserve of filtering to eliminate the noise. These measurements can be permanent measurements of gas pressure and temperature, rotational speed of a turbo-compressor shaft, and wealth probe voltage. Some permanent measurements, such as rapid cylinder pressure measurements, can also be used to calculate quantities describing combustion, for example a crankshaft angle for which 50% of the fuel fraction has burned for each cylinder, an instantaneous maximum pressure in the cylinder. each cylinder, a noise of combustion in each cylinder. Some measurements may be interrupted by purging or filling operations, such as the chemical species concentration of the flue gases, the opacity of the exhaust gases or the number of particles they contain, the instantaneous measurement of the fuel consumption. fuel. These measurements are performed by synchronous devices having a high availability rate. Each continuous measurement is performed over a given acquisition period, this duration being chosen so as to be able to determine at least one sigmoid function to model the signal. This time can be determined empirically, based on previous motor tests, and / or calculated during the process, so that the signal can be modeled by a sigmoid function. A mode of reevaluation of the duration of the process is detailed later in the description, in relation with step 8. In FIG. 8A, each measurement point calculated according to the present invention, and symbolized by a black square, is for example from the modeling of the signal on the first 60 seconds of acquisition with a single sigmoid function according to equation (I). An acquisition time d of 60 seconds is therefore sufficient in this example to allow the modeling of the signal by a sigmoid function according to the invention.

The acquisition time d can be fixed or variable from one signal acquisition to another, during an engine test or from one engine test to another. During the same motor test, during which the acquisition of several signals can be performed, a variable acquisition time can be adapted to the different quantities measured, some of which may be more reactive than others. This is the case for example for the richness of the gases obtained via an exhaust oxygen sensor, or the combustion noise which is calculated from the measurement performed by a cylinder pressure sensor. placed directly in the combustion chamber. These parameters are derived from measurements made with sensors located very close to the source they measure, and their response time is high, of the order of a second to a few seconds.

In the case where the duration d is fixed for the same motor test, this time d is preferably determined on the basis of the measured parameter which requires the longest acquisition time. For example, this operating parameter PF may be a pollutant such as unburned HC hydrocarbons which generally have the evolution of the slowest signal. Step 2: Modeling by at least one sigmoid function: qualification of the raw temporal sianal This stage aims at determining if the acquired signal can be processed by means of a modeling by at least one sigmoid function in order to calculate a representative value of the signal stabilized.

During this step, two criteria can be calculated that make it possible to qualify or not the signal: a first evolution criterion of evolution of the signal over the duration d, expressing the evolution of the signal over time. The first criterion evolves according to the following formula: evol = 1 average (raw signal on the first samples) mean (raw signal over the duration d) - a second criterion dyn on the dynamics of the raw signal with respect to its intrinsic variability , which amounts to evaluating an amplitude ratio of the signal to measurement noise. This second criterion is written in the following way: dyn = max (filtered signal) - min (filtered signal) standard deviation (raw signal - filtered signal) The filtering of the signal carried out for the calculation of this criterion can be different from that optionally carried out in step 4 described below. Advantageously, this formula makes it possible to avoid defining a noise value for each measurement. If the first criterion evol is less than a predetermined threshold value AA or if the second dynamic criterion dyn is lower than a predetermined threshold value of dynamic A, expressing the case of a signal which varies little or which is very noisy, then one considers that the signal is not comparable to a sigmoidal curve, and it does not continue with the following steps 3 to 8. In such a case, a value representative of the signal which does not evolve, and considered by Therefore stabilized as an alternative value ALT-PF representing the signal at infinite time, for example the median value of all samples of the acquisition.

FIGS. 6A and 6B are examples of unspecified signals based respectively on the evolution criterion of the evol signal (signal evolving too little) and the dynamic criterion of the dyn signal (too noisy signal). If the first and second criteria are satisfied (greater than the threshold values AA and A respectively), then the signal is qualified, i.e. it contains information that can be processed, and the following steps of the method can be performed. Step 3: Modeling ar at least one function if moid: value value calculation of the first asymptote) In this third step, the first time samples of the signal are used to definitively determine the value y o of the first asymptote. The value yo corresponds to the stable value recorded before changing the PR setting parameters and / or the motor power point. This parameter can be calculated whatever the sigmoid function (s) used to model the signal because it is necessarily part of the equation of the sigmoid function (s) modeling the signal. It is for example common to equations (I) and (II). Step 4 (optional): Modeling by at least one sigmoid function: filtering of the signal An optional step of filtering the acquired raw signal, represented in dashed lines in FIG. 5, can be carried out. Some signals may have significant measurement noise, which can be attenuated by smoothing the signal by applying a filter, which is chosen according to its sampling frequency. Preferably, the filter has a minimum phase shift effect, i.e. it produces a minimum time offset between the filtered signal and the raw signal. Among the filters that can be applied, the following filters, which are well known to those skilled in the art, can be cited in a nonlimiting manner, which can be chosen taking into account their degree of simplicity of implementation in a computer program and the computation time required: - a moving average, which has the advantage of being simple to implement and inexpensive in computing time. The moving average is preferably centered in order not to produce a phase shift of the signal; 3 0002 06 18 - first-order low-pass filtering, which has the advantage of being inexpensive in computing time. The use of such a filter preferably requires the compensation of the phase shift created; - a Butterworth filter. The use of such a filter preferably requires the compensation of the phase shift created; Step 5: Modeling with at least one sigmoid function: choice of the siomoid function (s) The purpose of this fifth step is to determine the sigmoid function (s) used for model the signal.

For that, the choice between the modeling by a sigmoid function according to the equation (I) or by two sigmoid functions according to the equation (II) is based on the computation of the temporal derivative of the signal, crude or filtered. Depending on whether the time derivative retains or changes sign, the modeling is chosen respectively according to equation (I) or (II). If the modeling by a sigmoid function according to the equation (I) is chosen, the unknowns to be determined are constituted by the four parameters ystep, b, c and t. If the modeling by two sigmoid functions according to equation (II) is chosen, it is the eight parameters v stepl, Ystep2, b1, C1, 1.11, b2, C2, ti2 that must be determined. Strand 6: Modeling ar a function so moid: calculation of arameters P s of the sipmoid function (s) chosen (s) This step aims to determine all the parameters of the function (s) not yet determined sigmoid (s) chosen in the previous step. For this purpose, a statistical regression using a least-squares method is used to determine the values of the four parameters or the eight parameters of the sigmoid function (s), the parameter yo having already been determined for the first time. step 3, from the raw or filtered signal. The least-squares minimization carried out aims at minimizing a criterion which is equal to the quadratic sum of the differences between the values of the acquired signal (raw or filtered) and the values of the signal modeled by the sigmoid function (s) ( s): E (sample of the raw or filtered signal - sample of the signal identifies samples When the values of the four (or eight) PFs parameters of the sigmoid function (s) evolve by converging the criterion to a minimum Optimal values are then obtained for these four (or eight) PFs parameters which allow an estimation of the acquired signal.In the case where the minimization process does not converge towards a solution (N conv in FIG. the acquisition time d and resume in step 1 and / or change the filter and resume in step 4 until convergence is achieved, or a traditional point capture Step 7: Determining a value représentati ve of the stabilized signal: calculation of the valet value of the second asymptote) This step makes it possible to determine the representative value of the stabilized signal for the operating parameter PF. For that, one calculates the value of the second asymptote of the graphical representation of the function (s) sigmoid (s), y. If the signal is modeled by at least one sigmoid function according to equations (I) or (II), the value of the second asymptote is calculated respectively according to equations (III) and (IV) below: Y = Y 0 ± Y step = YO + Y stepl Y step2 (IV) FIGS. 7A and 7B illustrate the calculation of the value of the second asymptote according to equations (III) and (IV). Knowing the four or eight parameters of the sigmoid function, we know the value of y. which represents the value of the stabilized temporal signal. Step 8: Determination of a representative value of the stabilized signal: qualification of the result From the values obtained for the parameters of the sigmoid function (s) PFS, the graphical representation of the function (s) is constructed. (s) sigmoid (s) modeling the signal, and calculating a correlation coefficient C ', between this estimated signal, and the acquired signal, raw or filtered. The correlation coefficient can be written according to the following formula: ## EQU1 ## The correlation coefficient varies between -1 (anti-correlated signals) and 1 (perfectly correlated signals). When it takes the null value, the signals are not correlated. Preferably, the correlation coefficient is calculated over the entire acquisition time of the signal (n then representing all the samples of the acquired signal over the duration d). Alternatively, the correlation coefficient is calculated on a time window located towards the end of the signal (n then representing the number of samples of the signal on this time window). If this correlation coefficient is high, for example greater than or equal to a predetermined correlation value B (for example 8 = 0.98), it is considered that the result is reliable, and the value of y is selected. (second asymptote) as a representative value of the stabilized signal CAL-PF. We can go to step 1 if another signal is acquired during the engine test. In the opposite case (Ccor <8), and if the correlation coefficient is not too far from 8, for example if the correlation coefficient is greater than (B - 0,1), we can proceed as in step 6 in the case of non-convergence, that is to say increasing the acquisition time d by going back to step 1 and continuing the acquisition of the signal, and / or modifying the signal filtering of the step 4, and repeat all or part of the intermediate steps until the case Ccor B If the result is not qualified after a number of reiterations of steps 1 and / or 4 or if the correlation coefficient is too far from B, you can take a traditional stitch. For example, it is possible initially to modify the filtering, and repeat steps 5 to 8. If Ccor is always less than B, then the acquisition time can be increased and / or the filtering of step 4 can be modified. Finally, if Ccor is always less than B, we can perform a traditional point capture.

Advantageously, the acquisition time d of the signal is increased at the end of this step of qualifying the result, in the case where the result is not qualified because of a correlation coefficient lower than the predetermined correlation value. 8, until the correlation coefficient is greater than the value B. For this, the order to continue the measurement is for example sent to the measurement instrumentation which provides the continuous measurement of the operating parameter, and repeats at least steps 1 and 4 to 8, a number of times sufficient to reach a correlation coefficient greater than the value B. It can be set a maximum number of times n max for which the duration d is increased to from this eighth step, for example n max equal to 2 times. The increase in duration d can be set by default to a given value d_more, for example d_more equal to 30 seconds. It will be ensured that the maximum duration d + (n_max * d_plus) is less than the average duration of a conventional point taking so as not to degrade the overall duration of the test sequence compared to a traditional approach.

The calculation time of steps 2 to 8 ensuring the processing of all operating parameters PF is short enough to allow this optional increase in the acquisition time d of the signal during the process. This calculation time is preferably less than the value of d_plus to allow the realization of the optional increase of d, provided that the signal acquisition unit is able to continue its processing independently of the unit. calculation which executes steps 2 to 8. This is the case of an automation system which combines at least these two units in the same set, and as described in Figure 10. Once determined all the values representative of stabilized signals acquired during the engine test sequence, relating to at least one operating parameter of the engine, all of these values are analyzed to characterize the operation of the engine. The evolution over time of each operating parameter, given by all the representative values during the engine test sequence, is recorded and / or displayed in a system comprising a display unit, a memory, and possibly a computing unit. The analysis can be performed automatically by the calculation unit, which compares, for example, the time evolution of the operating parameters of the engine under test with typical values, for example in order to ensure the operational safety of the engine or its organs in accordance with its specifications of use, or by a user of the process, who can make conclusions about the operation of the engine, for example in order to indicate the level of pollutants emitted, from the representative values determined and their evolution over time. Some engine test sequences are intended for example to adjust the operation of the engine at full load. During a sequence of this type, the engine is placed at a given speed and its maximum power as specified in the specifications, and the engine settings are varied by jumping from a starting setting. These jumps produce an evolution of the PF operating parameters of the engine, some of which are related to operational safety and are preferably analyzed. These parameters are for example the pressure of the gases inside the cylinder, which must not exceed a certain value at the risk of causing a buckling of a bièle, or the exhaust temperature, which must not exceed a certain value at risk of causing the turbine blades to melt together for a supercharged engine. Other test sequences are intended, for example, to regulate the operation of the engine in a pollution control zone, that is to say a zone corresponding to a range in engine speed / torque in which pollutant emissions are controlled in order to meet the requirements of the engine. regulatory constraints. During a sequence of this type, the engine is placed at a fixed speed and power representative of the conditions of use that the vehicle is likely to encounter following the driving profile of a pollutant homologation cycle. . From a start setting, the settings are varied by jump; these jumps will produce an evolution of the operating parameters PF of the engine and those that must be analyzed are related to the polluting emissions: for example the concentration of CO2 in the exhaust gases, which constitutes a global pollutant having an impact on the climate that is sought to be limited to meet the standards in force, or the concentration of unburned hydrocarbons (HC family) in these same exhaust gases, which constitute a local pollutant having an impact on the quality of the air. The unburned hydrocarbons must also be limited in order to comply with the standards in force (eg Euro5 emissions standards applying to new vehicles and until they reach 100000km under normal conditions of use).

Figure 8A shows the evolution of an operating parameter of an internal combustion engine, the concentration of CO2 in the exhaust gas, during a sequence of engine tests. During this test sequence, 189 rapid and simultaneous variations of the experimental plane settings on a fixed power point at (1750rpm, 2bar) are linked together. The total duration of the test sequence is 7:32, ie approximately 143 s per unit engine test. This test sequence is performed on a motor brake bench. The engine tested is a diesel engine (PSA DV6 Turbo-Diesel Injection Direct Common-Rail Diesel), powered with a standard fuel. The acquisition frequency of all the signals made is 1 Hz. The measured operating parameters are as follows: - pollutants: CO2, CO, HC, NOx, 02, recirculated CO2 at intake; - voltage delivered by a wealth probe; - instantaneous calculations: wealth "5gaz"; - temperatures: intake gas, exhaust gas; - pressures: intake gas, exhaust gas; - flow rates: fresh air on admission.

In FIG. 8A, the points symbolized by black squares are the values representative of 189 signals (189 points) acquired during the engine test sequence, calculated according to the method of the invention, with a modeling of each signal by a only sigmoid function according to equation (I) and an acquisition time d of 60s. By way of comparison, the diagram of FIG. 8A comprises points symbolized by gray circles representing the values calculated according to a traditional point taking, comprising a 30 second averaging phase for determining the measurement points. There is clearly a superposition of the two types of values, which reflects the good performance of the method according to the invention for obtaining in a short time representative values of the stabilized signals equivalent to those calculated in a traditional manner in a reduced time. The good coherence between the values calculated according to the invention and according to the prior art is also clearly visible in the diagram of FIG. 8B, which represents the residue, that is to say the difference between the value estimated according to FIG. invention and the average value calculated according to the prior art, between the two types of values as a function of the amount of CO2 measured. The set of residues has a divergent nozzle shape. FIG. 8C is a parity diagram, in which the CO2 values estimated according to the invention (CO2cAL on the ordinate) are represented as a function of the values calculated according to the prior art (CO2TRAD on the abscissa), is another way of representing the residues we see that the residue is weak, the points being close on the oblique line y = x.

FIGS. 9A, 9B and 9C represent, in the same manner as for FIGS. 8A to 8C, the evolution of the quantity of HC during the same engine test sequence, determined according to the invention and according to the prior art (FIG. 9A), as well as the performances of the process according to the invention (FIGS. 9B and 9C). Like the quantity of 002, we observe a very good correspondence between the values calculated according to a traditional point catch (gray crecles) and according to the invention (black squares). The residuals shown in Figure 9B are small, although a little more dispersed than for 002, and reflect a very good process performance for a measurement of HC that is generally more difficult to achieve than that of 002. Figure 9C show a very good correspondence, and therefore a low residue, between the HC values estimated according to the invention and the values of HC calculated according to a traditional point catch. The method according to the invention allows a significant time saving during the engine test compared to taking points in a conventional manner. This saving of time is estimated at about 50% of the time traditionally required to perform the engine test. In the examples presented in FIGS. 2, 3 and 4A, 4B, 4C, the duration of the engine test sequence is reduced by 5 hours and 15 minutes, or nearly 54%, considering a fixed acquisition duration of 60 seconds for each point followed by 25 seconds on average to obtain the values of asynchronous measuring systems, compared to a traditional test sequence that would have required 9:45 with an average acquisition time of 184 seconds for each point. In the examples presented in FIGS. 8A, 8B, 8C, 9A, 9B and 9C, the duration of the engine test sequence is reduced by 3 hours, ie by 41%, considering a fixed acquisition time of 60 seconds. for each point followed by 25 seconds on average to obtain the values of the asynchronous measurement systems, compared to a traditional test sequence that would have required 7:32 with an average acquisition time of 143 seconds for each point.

The method according to the invention makes it possible to predict the representative value of the stabilized signal, which makes it possible to anticipate the transition to the next continuous measurement during the test without having to wait for the complete stabilization of the signal. It is then saved the waiting time associated with at least the traditional averaging phase of the point taking. Such a saving of time not only allows a significant reduction in the total duration of the test, but also makes it possible to prevent the motor from reaching a prolonged thermomechanical limit, and in this way offers a better protection of the engine and its components. organs. The invention thus makes it possible to reduce the duration of the engine test while improving the operating safety of the engine. The time saved can help to speed up test speeds on engine benches or to obtain faster information on the operation of the engine in a vehicle in operation, but also help to limit engine wear and / or reduce or reduce maintenance operations on engines (engine oil change / replacement every 100 hours) and measuring equipment (biennial intervention for the exhaust gas analysis). This also has another virtuous effect which is the improvement of the quality of measurement insofar as the levels of repeatability and reproducibility of the measurements are reduced since the test sequences are linked more rapidly and the conditions are more homogeneous. The method according to the invention is robust and reliable in that, for each continuous measurement, a value representative of the stabilized signal is provided, in contrast to a traditional point pick where a greater variability of the result is observed when the present signal instability. The method of the invention also makes it possible to predict when a given value of the signal will be reached in the future. In the case where the signal representative of an operating parameter PF has a fairly slow dynamic, that is to say a significant transition time between the two asymptotes, and is of the same order as the acquisition time d, we will be able to estimate a stable value that is quite different from the last samples acquired after this time d. After having identified the parameters of the equation (I) for example, we can predict the moment in the future which separates the last acquired samples of a ((1 -1 Ytgt YO target value ytgt given on the path to yinfini trgt = --x ln - x ln Ystep Thus, it is possible either to correct the applied settings or to apply other settings in order to avoid a limit or to the main steps of the method according to the invention are preferably carried out automatically, advantageously using a computer program, so that it is possible to control other actuators during the available time interval. also disclosed is a computer program product for keeping in a memory of a unit of a processor or on a removable memory medium suitable for cooperating with said processor unit, the product-p program comprising instructions for implementing the method according to the invention. The invention also relates to a test device of an internal combustion engine (engine test bench) for performing engine tests in the context of a test sequence. FIG. 10 illustrates such a device, allowing the implementation of the method according to the invention. The device 100 according to the invention comprises: measuring instrumentation (102, 104, 105) for the continuous measurement of at least one operating parameter of said combustion engine 101 generating a raw time signal (Si, Bb, S2a) , S2b) during an engine test. This instrumentation may comprise one or more measurement probes (102, 104), for example fixed on the engine, on its members (for example an EGR valve or a turbo-compressor), or on sampling lines. The sampling lines are heated hoses, in order to prevent any condensation, which are responsible for conveying the exhaust gases, part of the flow of which is sucked up, by means of a pump, from the exhaust line to the various analyzers. , often more than a dozen, the bay analysis. This instrumentation also includes the conditioner (s) 105 (or converters) responsible for converting the signals acquired by the sensors (which can deliver for example an electric charge or a very low voltage) into a signal suitable for processing by the acquisition unit. (for example a voltage in a range from 0 to 10V); an automation system 106 for supervising the execution of the motor test and for displaying and / or recording the acquired signals with the measurement instrumentation (102, 104, 105). This supervisor 106 thus comprises a display device and / or a memory (recording unit) for displaying and / or recording the signals; a computer program according to the invention for determining a representative value of said stabilized signal; said automation system 106 capable of displaying and / or recording all of the representative values determined by the computer program, for their analysis to characterize the operation of said combustion engine 101; and said computer program being integrable with said automation system 106, or with said measurement instrumentation (102, 104, 105), or with a computer 107 separate from said automation system 106 and said measurement instrumentation (102, 104, 105), said computer 107 then being in communication with said automation system 106. The method according to the invention can be implemented in the context of motor tests carried out on motor benches, but also in the context of tests carried out on a vehicle in operation, that is to say circulating, comprising said engine. In the latter case, the vehicle comprises in particular an on-board instrumentation making it possible to carry out a continuous measurement of an operating parameter of the engine. The method according to the invention then provides access to information on the operation of the engine in real time or delayed, potentially used to automatically calibrate the engine and / or be used by the driver to modify its behavior and achieve operation 25 optimal vehicle (limitation of pollutant emissions, etc.). Advantageously, the present invention is used to calibrate the internal combustion engine. The internal combustion engine is calibrated from the analysis of all the representative values of the stabilized signals characterizing the operation of said engine. It is thus possible to modify engine adjustment parameters so as to optimize the operation of the engine, according to the predetermined specifications. Thus, it is for example possible to modify engine adjustment parameters related to the injection, the air mass in the cylinder, the exhaust gas recirculation (EGR), the supercharging, etc., according to operating parameters determined according to the invention, so as to limit the consumption of the engine and / or the polluting emissions. These modifications of the engine control parameters according to the information obtained from the method according to the invention can also be carried out outside the context of tests on engine benches, for example during the operation of the vehicle, as mentioned below. The method according to the invention can for example be used to determine the tens of parameters of a statistical model of behavior for each operating parameter PF of the engine, which can then make it possible to calibrate the motor. In this context, sequences of several hundred tests organized in experimental design are carried out. Power point jumps and motor adjustments from one point to another are obtained by searching for an optimal criterion (maximum of the determinant of the regressor matrix for a polynomial model or maximum of the minimum interpoint distances). for an interpolatory type model). This type of sequence, performed in a conventional manner with traditional scans, is very expensive in time (several tens of hours, sometimes 2 days continuously). The use of the method according to the invention, which can be used to halve the duration of the sequence, is therefore particularly well suited for this application. The following steps can be performed to determine the behavior models of the operating parameters and to characterize the engine, for example in view of its calibration: 1.1 list the actuators of the engine to be driven; 1.2 define the minimum and maximum limits of variation of the setting of each actuator. The method according to the invention can be implemented at this stage; 1.3 identify interdependencies between settings that can not vary independently of each other (eg fresh air intake and boost pressure). The method according to the invention can be implemented at this stage; 1.4 calculate the experimental plan; 1.5 perform the test sequence more quickly by the method according to the invention by following the combinations of settings listed in the plan; 1.6 calibrating the parameters of each statistical model of engine operating parameters by using the final values estimated by the method according to the invention; 1.7 optimize the statistical models and determine the actuator settings that make it possible to satisfy the operating constraints of the engine; 1.8 gather all of these optimum settings in a table (one table per setting) that will be entered in the electronic engine computer that controls the engine fitted to each vehicle produced in series; all of these tables can be used for engine calibration.

The invention can also be used to exhaustively characterize the evolution of each operating parameter PF of the engine in its range of use in speed and power. For this purpose, for example, increasing variations are made by actual (or indicated) average torque jump (or pressure) over a growing series of engine speed levels. This type of test sequence lasts about ten hours but is repeated several times during a study; that's why halving the execution time is interesting. The steps for this type of test may be the following: 2.1 load into the calculator a set of optimal tables (see step 1.8 previously described); 2.2 perform the test sequence more quickly thanks to the method according to the invention starting at low speed and low torque and then achieving iso-steady torque jumps to the maximum torque and then increasing the speed after having reduced the couple to a minimum; 2.3 enter in a table indexed in speed and torque levels the values of the operating parameters of the engine using the final values estimated by the method according to the invention; all of these tables form an engine ID card when set with these optimal settings tables. The method according to the invention can also be used to test the engine once it is put on the market, in order to evaluate its proper operation. Indeed, some programs that run in the electronic computer are intended to verify that certain organs still fulfill their role in the cleanup of exhaust gases (eg the EGR valve or the particle filter). These programs detect when the vehicle in use is part of a logical sequence of states in order to make a reading of an operating parameter of the engine (eg a position copy or a pressure difference). When this reading does not comply with a pre-recorded value, depending on the severity of the malfunction of the pollution control system, either a fault is recorded in the computer or a light can be lit on the dashboard. In the latter case, the user is then alerted to a malfunction that requires a return to the workshop.

The invention can facilitate the achievement of the survey and reduce the risk of incorrectly detecting a malfunction.

Claims (16)

REVENDICATIONS1. A method of characterizing the operation of an internal combustion engine from a motor test sequence, each engine test comprising modifying at least one adjustment parameter of the engine PR and / or the power point of the engine, comprising the following steps: a) for each motor test: I. at least one signal is obtained from the continuous measurement of an operating parameter PR of said motor over an acquisition period of the acquisition duration d being chosen so as to be able to determine at least one sigmoid function for modeling said at least one signal; ii. said at least one sigmoid function modeling said at least one signal; iii. a representative value of said at least one stabilized signal is deduced from said at least one sigmoid function; b) analyzing all of said representative values of the stabilized signals to characterize the operation of said internal combustion engine. 2. Method according to claim 1, wherein at least one adjusting parameter PR and / or the power point of the motor is varied rapidly, preferably between 0 and 5 seconds, after the triggering of the signal acquisition. 3. Method according to one of the preceding claims, wherein the continuous measurement is performed at a sampling frequency between 1 Hz and 10 Hz. 4. Method according to one of the preceding claims, wherein said at least one sigmoid function is a Gompertz function. 5. Method according to one of the preceding claims, wherein the graphical representation of said at least one sigmoid function comprises a first horizontal asymptote whose value is determined at an initial moment 1'0 of the signal and a second horizontal asymptote whose value is determined at an infinite moment t .. 6. Method according to claim 5, in which the signal is modeled by a sigmoid function written according to the following equation (I): y (t) = yo + y step x exp (-bx exp (-cx (t -))) (I) comprising the five parameters to be determined yo, Ystep, b, c and yo being the value of the operating parameter PF at an initial time t = t0; ystep being the difference between the value of y (t) at infinite time t = t., and 1.0; b and c being positive constants, with b characterizing the shift of the curve over time and characterizing the slope of the curve; ti being being a positive time constant 7. The method of claim 5, wherein the signal is modeled by two sigmoid functions. The method of claim 7, wherein the two sigmoid functions are combined under the following equation (II): y (t) = yo + y step x exp (-b1 x exp (-c1 x (t-t11) )) + Ystep2 x exp (-b2 x exp (-c2 x (t-ti2))) (Il) including the nine parameters to be determined y0, 0,, V step1, Ystep2, b1, C1, t11, b2, C2 , t12; YO being the value of the operating parameter PF at an initial moment t = t0; Ystepl being the difference between the value of y (t) at an infinite time t = t ..> and that at t, if a single first thermomechanical and / or thermo-chemical phenomenon existed, and representing the variation of the signal related to this first phenomenon; Ystep2 being the difference between the value of y (t) at infinite time t = t .., if only the first phenomenon existed and the value of y (t) at infinite time t = t., If only one second thermomechanical phenomenon and / or thermo-chemical antagonist at the first phenomenon existed, Ystep2 representing the variation of the signal related to this second phenomenon; Ystep1 and Ystep2 being of opposite signs; b1 and c1 being positive constants, with b1 characterizing the shift of the first S-curve in time, and c1 characterizing the slope of the first S-curve; being a positive time constant relative to the inertia of the first phenomenon! b2 and c2 being positive constants, with b2 characterizing the shift of the second S-curve in time, and c2 characterizing the slope of the second S-curve; ti2 being a positive time constant relative to the inertia of the second phenomenon. 9. The method of claim 6 or 8, wherein the step ii) of determining said at least one sigmoid function modeling the signal comprises the following steps: the signal having a dynamic criterion or an evolution criterion is selected; temporal greater than or equal to a predetermined dynamic threshold value or a predetermined evolution threshold value (QUAL-S); the parameter yo corresponding to the value of the first horizontal asymptote (Val-AS1) is determined; the time derivative of the signal (CH-FS) is calculated; the sigmoid function is defined according to equation (I) or the two sigmoid functions according to equation (II) according to whether the sign of the time derivative remains the same or changes (CH-FS); the other parameters of said sigmoid function (s) are calculated in such a way as to minimize, in the least squares sense, the difference between the acquired signal and the estimated signal according to the said sigmoid function (CAL-PFs); and wherein step iii) of determining the representative value of the stabilized signal comprises the following steps: determining the value of the second asymptote y (Val-AS2); a correlation coefficient is calculated between the acquired signal and the sigmoid functions and y is selected as a representative value of the stabilized signal if the correlation coefficient is greater than a predetermined correlation value (QUAL-corr, CAL). -PF). 10. The method of claim 9, wherein increasing the acquisition time d after step iii) of determining the representative value of the stabilized signal so that the correlation coefficient is greater than one. predetermined correlation value. 11. Method according to one of the preceding claims, further comprising a step of filtering the signal (FIL) during step ii) of determining the sigmoid function modeling the signal, said filtering comprising a filter preferably selected in the list of following filters: moving average, Butterworth filter, first-order low-pass filter. 12. Method according to one of the preceding claims, wherein the engine test is performed on a test engine or on a running vehicle. 13. Method according to one of the preceding claims, wherein the internal combustion engine is calibrated from the analysis of all the representative values of the stabilized signals characterizing the operation of said engine. 14. Method according to one of the preceding claims, used to characterize the operation of a compression ignition engine. 15. Computer program product for storing in a memory of a unit of a processor or on a removable memory medium suitable for cooperating with said processor unit, the program product comprising instructions for implementing the method according to the present invention. any one of claims 1 to 14. 16. A test device (100) for an internal combustion engine (101) for performing an engine test sequence, comprising: - measurement instrumentation (102, 104, 105) for continuous measurement of at least an operating parameter PF of said motor (101) generating a raw time signal (S1a, S1b, S2a, S2b) during a motor test; an automation system (106) for supervising the execution of the motor test and for displaying and / or recording the acquired signals with the measuring instrumentation (102, 104, 105); a computer program according to claim 15 for determining a representative value of said stabilized signal; said automation system (106) being able to display and / or record all of the representative values determined by the computer program, for their analysis to characterize the operation of said combustion engine (101); and said computer program being integrable with said automation system (106) or said measurement instrumentation (102, 104, 105), or a computer (107) separate from said automation system (106) and said instrumentation measuring device (102, 104, 105), said computer (107) being in communication with said automation system (106).