METHOD FOR DIAGNOSING A SENSOR AMONG A SET OF SENSORS

Abstract

Method for diagnosing a sensor (Ti) among a set of sensors (Ti), characterized in that this method executes in order and successively:- a start step (10) inhibiting the sensors (Ti) in the invalid state (Tin),- a step (20) determining the average value (Tmoy) of the values of the sensors (Ti) in the valid state (Tiv),- a step (30) determining, in absolute value, each deviation (∆Ti) between the value of each sensor (Ti) in the valid state (Tiv) and the average value (Tmoy), and determining the maximum deviation (∆Tmax) among all of these deviations (∆Ti),- a step (40) of comparing each of the deviations (∆Ti) with a threshold value (S) common to all of the deviations (∆Ti), and if one of the deviations (∆Ti) is greater than this threshold (S) and this deviation (∆Ti) is the maximum deviation (∆Tmax) then the sensor (Ti) associated with this deviation (∆Ti) is diagnosed as being in the invalid state (Tin). Figure 1.

Description

Title of the invention: METHOD FOR DIAGNOSING A SENSOR AMONG A SET OF SENSORS

[0001] In the field of motor vehicles, and in particular vehicles comprising a fuel cell supplied with fuel by hydrogen tanks, knowledge of the temperature of the hydrogen stored in these tanks, at any time, is an essential link for the operational safety of the vehicle and its occupants.

[0002] Diagnostics of these temperature sensors have also been developed to determine a valid or invalid state of each of these sensors, the invalid state meaning, for example, a failure of the temperature sensor as such, or of any part of the acquisition chain of the measured value of this sensor.

[0003] Thus, if there are several sensors, it is known to those skilled in the art to compare the values of two sensors with each other, and to establish a diagnosis based on the deviation observed with respect to thresholds calibrated and stored in data tables, each threshold being a function of each pair of sensors compared and of the temperature measured among other things. Unfortunately this diagnosis depends on this calibration, and if a calibration error has occurred the diagnosis itself becomes fallible. This risk of incorrect diagnosis is even higher when there are more than two sensors, in particular three, four, or more: the number of necessary comparisons of sensors two by two becomes exponential as a function of the number of sensors, as do the calibration and the tables and the risk of error increases considerably. In addition to the fact that this diagnosis quickly becomes complex and consumes memory and computing resources, the risk of error is high.Furthermore, if there are only two sensors, this diagnosis cannot discriminate which one would be invalid.

[0004] The aim of the invention is to remedy this problem. To this end, the invention relates to a method for diagnosing a sensor from among a set of sensors, this method being such that it executes in order and successively: - an initial step performing an acquisition of a valid and invalid state of each of the sensors from a memory storing each of these states, or setting each of the states to the valid state in said memory, - a start step inhibiting the sensors in the invalid state to retain only the values of the sensors in the valid state, - a step determining the average value of the values of the sensors in the valid state, - a step determining, in absolute value, each deviation between the value of each sensor in the valid state and the average value, and determining the maximum deviation among all of these differences, - a step of comparing each of the deviations to a threshold value common to all of the deviations, and if one of the deviations is greater than this threshold and this deviation is the maximum deviation then the sensor associated with this deviation is diagnosed as being in the invalid state, all the other sensor(s) being diagnosed as being in the valid state, - a step of recording in the memory the invalid state of the sensor associated with this deviation if this sensor has been diagnosed as invalid, and whatever the valid or invalid states of each of the sensors of the set of sensors the method executes a return to the start step until there remains only one sensor in the valid state.

[0005] A sensor in the invalid state is a sensor likely to be defective or at least faulty. As described previously, for the purposes of the method which is an iterative method, the value of the invalid sensor is ignored (no longer taken into account) for the following iteration. This method is only a diagnostic method; it does not cover possible corrective actions which could be envisaged by those skilled in the art, such as for example a verification of the sensor diagnosed as invalid by another redundant method, or the neutralization of actions or securing of organs whose operation depends on the invalid sensor.

[0006] Throughout the text of this document, the sensor and its measured value will be considered equivalent, and will bear the same sign.

[0007] Thus this method allows for very rapid convergence since at each iteration it is possible to diagnose a sensor in the invalid state, and, i being the number of sensors, if i-1 sensors are in the invalid state it will only take i-1 iterations to diagnose these i-1 sensors, which is a much lower number of iterations compared to the methods of the prior art. The risk of error is therefore lower as is the load on a computer capable of implementing this method. Of course, if there is only one sensor left in the valid state this method no longer makes sense and ends, however the fact of considering the condition "and" on "and if one of the deviations is greater than this threshold AND that this deviation is the maximum deviation" makes it possible to ensure that this last sensor is potentially the least faulty.Of course, if there is only one sensor left in a valid state (or the least faulty) the person skilled in the art knows that he will have urgent actions to take, but that is not the object of this invention.

[0008] According to one embodiment of the invention, the threshold is a constant value, which allows a simplification of the method.

[0009] According to an alternative embodiment of the invention, the threshold is a function of a model of one or more measured values of the sensors. This model models, for example, the measurement imprecision of the sensors Ti as such, in particular as a function of the measured value Ti such as a temperature for example, but this model can be more complex by integrating a dynamic inaccuracy depending on the gradients of the measured values, or ambient parameters to the sensors or the sensor electronics such as a humidity level, a temperature, among others. We will distinguish the electronics of the sensor from the sensitive element of the sensor, the electronics processing the signal from the sensitive element, the two together forming the sensor.

[0010] According to one embodiment of the invention, the threshold is a function of the average value.

[0011] According to one embodiment of the invention, each sensor of the set of sensors measures a physical value of the same nature.

[0012] Throughout the text of this document, a physical value of the same nature will be understood to mean values of the same dimension, for example a set of temperatures, or pressures / forces / mass, or current / voltage, or lengths, or time, or quantity of matter, or luminous intensity.

[0013] The method is of course more precise if each sensor in the set of sensors measures values likely to change in the same way, i.e. in the same direction, ideally in an identical manner, but this is not obligatory: - these developments can be delayed in time, it would then be advantageous to apply the process during a period when the measured temperatures are stable, or failing that, increase the threshold value, which makes the process less precise than waiting for the measured values to stabilize, - these developments may have differences between them, and in this case the threshold will have to take this into account by taking a high value, which makes the process less precise than if the developments were identical.

[0014] The invention also relates to a calculator comprising the means of acquisition, processing by software instructions stored in a memory as well as the control means required for implementing the method as previously described.

[0015] The invention also relates to a motor vehicle comprising a set of sensors and the computer as previously described.

[0016] According to one embodiment of the invention, this vehicle comprises a set of tanks each comprising a sensor from the set of sensors.

[0017] According to one embodiment of the invention, this vehicle comprises a powertrain and a fuel cell capable of supplying electrical energy to the powertrain, the tanks being hydrogen tanks for the fuel cell, and all of the sensors being temperature sensors for the hydrogen contained.

[0018] According to one embodiment of the invention, the set of sensors is a set of sensors all identical.

[0019] Other features and advantages will appear on reading the following description of a particular, non-limiting embodiment of the invention, made with reference to the single [Fig.l] in which:

[0020] [Fig.l]: represents a flowchart for executing the diagnostic method according to the invention.

[0021] [Fig.l] discloses a flowchart of a diagnostic method according to the invention. This method for diagnosing a sensor Ti among a set of sensors Ti, executes in order and successively: - an initial step 00 executing an acquisition of a valid state Tiv and an invalid state Tin of each of the sensors Ti from a memory storing each of these states, or setting each of the states to the valid state Tiv in said memory, - a start step 10 inhibiting the Ti sensors in the invalid state Tin to retain only the values of the Ti sensors in the valid state Tiv, - a step 20 determining the average value Tmoy of the values of the sensors Ti in the valid state Tiv, - a step 30 determining, in absolute value, each deviation ATi between the value of each sensor Ti in the valid state Tiv and the average value Tmoy, and determining the maximum deviation ATmax among all of these deviations ATi, - a step 40 of comparing each of the deviations ATi to a threshold value S common to all the deviations ATi, and if one of the deviations ATi is greater than this threshold S and this deviation ATi is the maximum deviation ATmax then the sensor Ti associated with this deviation ATi is diagnosed as being in the invalid state Tin, all the other sensor(s) Ti being diagnosed as being in the valid state Tiv, - a step 50 of recording in the memory the invalid state Tin of the sensor associated with this deviation ATi if this sensor Ti has been diagnosed as invalid, and whatever the valid or invalid states of each of the sensors Ti of the set of sensors, the method executes a return to the start step 10 until there remains only one sensor Ti in the valid state Tiv.

[0022] The index i of the signs designates an integer ranging from 1 for the smallest, to the total number of sensors Ti of the set of sensors Ti considered.

[0023] The initial step 00 allows the method to begin, but since it is an iterative method, it can be processed in two different ways: either the last invalid states Tin, from a previous execution of the method, are stored and reused as data during the next execution, or they are not stored and the method begins the next step (the start step 10) with all the sensors Ti in the valid state Tiv. In both cases, the iteration of the method is very fast. For example, for a motor vehicle, the first execution of the method took place during a previous run, then the vehicle is parked for a more or less long period and its computers are put into standby mode. For driving following this standby mode, the vehicle is restarted and the computers can execute a second execution of the process, which will therefore begin with the initial step 00.

[0024] The start step 10 inhibits the sensors Ti in the invalid state Tin to retain only the values of the sensors Ti in the valid state Tiv, thus for this method and the current iteration, only the values of the sensors in the valid state Tiv are taken into account, in particular for the determination of the average value Tmoy of step 20.

[0025] Step 40 combines two conditions: - if one of the ATi deviations is greater than this threshold S, and - if this deviation ATi is the maximum deviation ATmax, then the sensor Ti associated with this deviation ATi is diagnosed as being in the invalid state Tin. The sensor associated Ti with the deviation ATi is of course the sensor Ti which is compared to the average Tmoy and whose deviation ATi satisfies the two conditions above. Thus this method only diagnoses one sensor Ti per iteration, this sensor corresponding to the one giving the maximum deviation ATmax, the other sensors being considered to be in the valid state Tiv for the next iterations. Thus, in the case of several faulty sensors Ti, this method makes it possible to diagnose or discard first the most faulty sensor Ti.But in addition, and for the process to be precise so as not to generate false diagnoses, it is useful to first discard the value of the sensor Ti associated with the maximum deviation ATmax which exceeds the threshold S because this value is potentially the one which has the greatest influence on the determination of the average Tmoy and therefore on all the deviations ATi of all the other sensors Ti, some of which could also exceed the threshold S even though they are not faulty. This iteration, following step 50 which returns to the start step 10, makes it possible to discard the invalid value Tin and the next iteration of the process recalculates a new average Tmoy and so on.

[0026] Thus, for each invalid state diagnosis Tin, the average Tmoy changes discretely, and allows the diagnosis to continue on the other sensors Ti whose state is valid Tiv.

[0027] The threshold S is common to all the ATi deviations.

[0028] In one embodiment of the invention, the threshold S is a constant value, but this is not obligatory and, for example, the threshold S is a function of a model of one or more measured values of the sensors Ti. A simple example to illustrate this, independently of the device in which the sensors Ti are arranged, is the taking into account of the imprecision of the sensors Ti as a function of the measured value: it is known that this imprecision varies, it is in particular low over a given range of use, and increasing as one moves away from it. This type of variation can be taken into account in the determination of the threshold S, the value of which can then increase with the growth of imprecision. For example, the threshold S is a function of the average value Tmoy.

[0029] For example, each sensor Ti of the set of sensors Ti measures a physical value of the same nature, which leads to a simple example of implementation of this method, below.

[0030] It will be noted that this method can be executed during predefined periods, during which the measured values Tin, Tiv are stable: For example, after a long stop of the vehicle, during its parking, or before or during the starting of the vehicle over a period for example of a few seconds to a few minutes, for example 2 minutes, so as not to have a determination of the threshold S that is too complex and therefore a source of approximations.

[0031] This method is for example implemented by a computer comprising the means of acquisition, processing by software instructions stored in a memory as well as the control means required for implementing the method.

[0032] This method applies for example to a motor vehicle comprising the set of sensors Ti and the computer.

[0033] This vehicle comprises for example a set of tanks each comprising a Ti sensor from the set of Ti sensors. Advantageously the tanks are identical, for example 3 or 4 in number, or even more, the Ti sensors are identical and are temperature sensors of the contents in the tanks. Ideally these tanks each comprise a single Ti temperature sensor, but of course it is not excluded to have two or more Ti sensors per tank.

[0034] This vehicle comprises, for example, a powertrain and a fuel cell capable of supplying electrical energy to the powertrain, the tanks being hydrogen tanks for the fuel cell, and all of the sensors Ti being temperature sensors for the hydrogen contained in these tanks.

[0035] This common threshold S can for example be attributed to a given range of temperature values Ti, and beyond this range either the process is not executed, or the threshold S is adapted and takes other values (progressively or in stages).

[0036] If these reservoirs are identical, but for example one of them is more exposed to a source of heat or cold than the others, this difference will have to be integrated into the determination of the threshold S which will be increased, and in doing so the diagnosis will be less precise and will detect an invalid state later. Conversely, if each value or temperature Ti evolves simultaneously in the same way, and at very close or even identical values, the value of the threshold S can be reduced and the method will gain in precision, detecting a sensor in an invalid state very quickly.

[0037] It should be noted that it is not mandatory that each sensor Ti of the set of sensors Ti measures a physical value of the same nature. For example, the values can be linked by a physical relationship, such as the ideal gas law. Thus, among the Ti sensors, a majority of them will be able to measure a temperature while others will measure a pressure, especially if the volume of gas is constant, such as hydrogen tanks in prolonged storage situations: the temperature of the gas can be assimilated to the ambient temperature, and for a known quantity of gas (measurement of the mass, for example), a measurement of the pressure then indirectly gives a measurement of the temperature of the gas. It is of course obvious that it will be this indirect temperature that will be used to determine the average value Tmoy.

[0038] Thus for the simplest example, the vehicle comprises a set of tanks each comprising a sensor Ti from the set of sensors Ti, advantageously the tanks are identical, for example 3 or 4 in number, or even more, the sensors Ti are identical and are temperature sensors of the contents in the tanks, these tanks each comprise a single temperature sensor Ti, and the vehicle comprises a powertrain and a fuel cell suitable for supplying electrical energy to the powertrain, the tanks being hydrogen tanks for the fuel cell, and the set of sensors Ti are temperature sensors of the hydrogen contained in these tanks: The threshold S is of constant value for all the deviations ATi. It can be for example of the order of 5°C for this type of application, or between 2 and 10°C depending on the accuracy of the sensors and therefore the technology of the sensors.Alternatively, this threshold S is for example a percentage of the average temperature Tmoy and can therefore vary from one iteration to another automatically, increasing the diagnostic accuracy of the process as the sensors Ti in the invalid state are removed, and also taking into account an increasing inaccuracy of the sensors Ti as a function of the rise in the measured temperature represented by the average temperature Tmoy, for example a threshold S whose value is set from 1% to 5% of the average temperature Tmoy according to the characteristics of the sensors Ti. As a further variant, this threshold S can be determined by a one-dimensional table having temperature ranges as input.

[0039] This example of a fuel cell is an ideal example for this method, since it is desired that the pressure in each of the tanks is identical, in other words the flow rate of gas entering or leaving each of the tanks is identical. Thus the evolution of the temperature of each tank is identical and under these conditions the method according to the invention has the best possible precision, thus the threshold S can be set at 5°C, or even less, for example 3°C. Conversely, if one of these tanks is exposed to thermal conditions different from the others, for example thermally less insulated, or more exposed to a cooling fluid, a temperature difference due to this arrangement may appear between the measured values Ti without the sensors Ti being faulty - invalid. This difference will influence the diagnosis and may lead to a false detection of an invalid sensor and to avoid this, the value of the threshold S will be increased in this case, making the process less precise at the same time.

Claims

Claims

1. Method for diagnosing a sensor (Ti) among a set of sensors (Ti), characterized in that this method executes in order and successively: - an initial step (00) executing an acquisition of a valid (Tiv) and invalid (Tin) state of each of the sensors (Ti) from a memory storing each of these states, or putting each of the states in the valid state (Tiv) in said memory, - a start step (10) inhibiting the sensors (Ti) in the invalid state (Tin) to retain only the values of the sensors (Ti) in the valid state (Tiv), - a step (20) determining the average value (Tmoy) of the values of the sensors (Ti) in the valid state (Tiv), - a step (30) determining, in absolute value, each deviation (ATi) between the value of each sensor (Ti) in the valid state (Tiv) and the average value (Tmoy), and determining the maximum deviation (ATmax) among all of these deviations (ATi),- a step (40) of comparing each of the deviations (ATi) with a threshold value (S) common to all of the deviations (ATi), and if one of the deviations (ATi) is greater than this threshold (S) and this deviation (ATi) is the maximum deviation (ATmax) then the sensor (Ti) associated with this deviation (ATi) is diagnosed as being in the invalid state (Tin), all the other sensor(s) (Ti) being diagnosed as being in the valid state (Tiv), - a step (50) of recording in the memory the invalid state (Tin) of the sensor associated with this deviation (ATi) if this sensor (Ti) has been diagnosed as invalid, and whatever the valid or invalid states of each of the sensors (Ti) of the set of sensors, the method executes a return to the start step (10) until there remains only one sensor (Ti) in the valid state (Tiv).,

2. Method according to one of the preceding claims, the threshold (S) being a constant value.

3. Method according to claim 1, the threshold (S) being a function of a model of one or more measured values of the sensors (Ti).

4. Method according to claim 3, the threshold (S) being a function of the average value (Tmoy).

5. Method according to one of the preceding claims, each sensor (Ti) of the set of sensors (Ti) measuring a physical value of the same nature.

6. Calculator, characterized in that it comprises the means of acquisition, of processing by software instructions stored in a memory as well as the control means required for implementing the method according to any one of the preceding claims.

7. Motor vehicle comprising a set of sensors (Ti), characterized in that it comprises a computer according to the preceding claim.

8. Vehicle according to the preceding claim, comprising a set of tanks each comprising a sensor (Ti) from the set of sensors (Ti).

9. Vehicle according to the preceding claim, comprising a powertrain and a fuel cell capable of supplying electrical energy to the powertrain, the tanks being hydrogen tanks for the fuel cell, and the set of sensors (Ti) being temperature sensors for the hydrogen contained.

10. Vehicle according to one of claims 7 to 9, the set of sensors (Ti) being a set of sensors which are all identical.

Images