Methods for analyzing and evaluating measured values ​​of a test system

Abstract

Method for analyzing and evaluating measured values ​​of a test system designed as an engine test bench in which internal combustion engines are examined, wherein a test specimen is monitored during a test run by means of a plurality of sensor means and each of the sensor means provides measured values ​​of at least one measured quantity to an evaluation device for further processing, characterized in that a logic tree is generated for a plausibility check of the measured values ​​of the measured quantities, by means of which a plausibility is determined according to predefined rules by forming and using a presumably true value for each of the measured quantities.

Description

[0001] The present invention relates to a method for analyzing and evaluating measured values ​​of a test system, which is designed as an engine test bench in which internal combustion engines are examined, wherein a test specimen is monitored during a test run by means of a plurality of sensor means and each of the sensor means provides measured values ​​of at least one measured quantity to an evaluation device for further processing.

[0002] Measuring and testing devices, which can be used for research and development purposes in automotive engineering, for example, are known in various designs. When using such measuring and testing devices—for example, for investigating an internal combustion engine and a motor vehicle's powertrain—even relatively short test runs generate a very large volume of data, which must be stored, processed, and evaluated. Key objectives of software-based measurement data evaluation include, in particular, the assessment of the test runs and the test specimen itself. The test runs and the evaluation of the resulting measured values ​​can preferably be automated to optimize the development process.The quality of the measurement results depends crucially on whether the measuring and testing equipment used for the test runs functions flawlessly even over extended test periods. In this context, it is particularly important that the measured values ​​acquired by the sensors are plausible. Implausible measured values ​​from the outset, which can result, for example, from a malfunctioning sensor, render the results of the corresponding test runs of limited use. Repeating the test runs involves additional time, personnel, and costs. Therefore, it is desirable to ensure the plausibility of the measured values ​​through suitable plausibility checks and to detect implausible deviations early on – preferably during the test run itself.

[0003] From EP 1 246 034 A2, a method of the type mentioned above for analyzing and evaluating measured values ​​from a test system designed as an engine test bench for examining internal combustion engines is known. For plausibility checks, at least one plausibility node is coupled to at least one measuring channel of the measuring and testing device. The plausibility nodes used in this method, in combination with the type and number of available measuring channels, provide a value for the confidence level of at least one of the measured values ​​from the measuring and testing device. In their most general form, the plausibility nodes are rules that allow a statement to be made about the plausibility of the measured values ​​provided by a measuring channel linked to the plausibility node. The plausibility check rules used are based on technical-physical models.

[0004] German patent DE 10 2011 081 240 A1 discloses a method for testing a motor vehicle's braking system, which can be controlled in a so-called "brake-by-wire" mode both by the driver and independently of the driver. In this method, which is performed while the motor vehicle is in operation, the following measured variables are redundantly acquired using suitable sensors: pedal pressure of a master brake cylinder, travel of a brake pedal, pressure of a pressure supply device of the braking system, and position of a piston of the pressure supply device. The acquired measurement signals of these variables are compared in a sensor plausibility block. If both signals are within a defined tolerance band, a plausible measured value is output for the corresponding measured variable.

[0005] German patent DE 10 2013 111 218 A1 discloses a method for controlling, regulating, diagnosing, and / or monitoring a component of compressed air generation, compressed air treatment, compressed air storage, and / or compressed air distribution. This method allows for the derivation of plausibility criteria for real measurement / sensor values ​​from structural models to detect malfunctions or defects, and for verifying compliance with these plausibility criteria for real, current, or historical measurement / sensor values. For example, if the actual values ​​deviate from the values ​​considered plausible based on the models by more than a predefined threshold, this can be assessed as a malfunction, and a corresponding error message can be issued.Such plausibility criteria may in particular include the comparison of temperatures and / or pressures at measuring points that are arranged upstream or downstream of each other in flow paths of media (especially compressed air, cooling air, cooling water), whereby systematic increases or decreases in temperatures and / or pressures occur or are to be expected between the measuring points during trouble-free operation of the components.

[0006] The present invention aims to provide a method of the type mentioned above for the analysis and evaluation of measured values ​​of a test system, which enables the plausibility of the measured values ​​in an alternative way and in particular enables an increase in the expressiveness, a minimization of the parameterization effort, an extension of the valid operating range and a minimization of the complexity.

[0007] The solution to this problem is provided by a generic method with the features of the characterizing part of claim 1. The dependent claims relate to advantageous embodiments of the invention.

[0008] An inventive method for analyzing and evaluating measured values ​​from a test system configured as an engine test bench for examining internal combustion engines is characterized by the creation of a logic tree for plausibility checks. This logic tree, based on predefined rules, determines the plausibility of each measured value by generating and using a presumably true value. The inventive method provides a novel approach for verifying the plausibility of measured values ​​of various parameters acquired during a test run using sensors. The method is particularly distinguished by increased reliability, minimized parameterization effort, an expanded valid operating range, and reduced complexity.The method can advantageously be performed during the test run ("online"), so that potential errors in the measuring and testing equipment and / or the test specimen can be detected immediately during the test run. The method according to the invention thus makes it possible to make statements about the plausibility of the measured values ​​even while a test run is being carried out. The method is intended for use in engine test benches where internal combustion engines are examined. The test runs are carried out when all calibrations and functional tests integrated into the process have been completed and all simple tests indicate that the measuring and testing equipment appears to be working correctly.

[0009] To assess the significance of the signals for each other, two aspects in particular are taken into account: 1. Assessment of the relationship and correlation 2. Assessment of the reliability of a signal

[0010] To assess robustness, the accuracy of the sensor, its repeatability in conjunction with the device under test, and the failure probability ("robustness") of the signal are evaluated. These aspects are considered when constructing the logic tree, ensuring that accurate and robust signals are used for plausibility checks rather than less accurate and robust ones. The validated signals are then used to validate further signals, creating a cumulative logic of maximum robustness.

[0011] In a preferred embodiment, it is proposed that redundant measurements be performed using a number of redundant sensors to verify the plausibility of at least some measured variables of the test system. The redundant acquisition of measured values ​​of a specific variable using at least two redundant sensors enables simple plausibility verification and determination of the likely true value of that variable. This is particularly suitable for variables where the associated sensors provide consistently reproducible and reliable measurement signals. Examples of such variables include the rotational speed and fuel consumption (mass flow rate) of an internal combustion engine.Preferably, the presumably true value of a measured quantity is set as implausible if the presumably true value of this measured quantity lies outside an accuracy interval of a reference value measured with the at least one redundant (additional) sensor means.

[0012] In a further advantageous embodiment, it is proposed that, to verify the plausibility of the measured values ​​of at least some parameters of the test system, the acquired measured values ​​are compared with the results of at least one information linking equation for that parameter. In this context, it is preferred that mixture equations be used as information linking equations instead of balance equations. Balance equations often have the disadvantage of containing undesirable circular references.

[0013] In a preferred further development, it is possible to compare the recorded measured values ​​of at least some parameters of the test system with the results of at least one mathematical calculation model for these parameters to verify their plausibility. Preferably, the input parameters of the at least one mathematical calculation model are already plausible values ​​of some parameters that have been verified through comparative measurements or information linking equations and are likely to be true.

[0014] For each measured variable, at least one reference value and one plausibility interval must be established. This plausibility interval is derived from measurements with a reference device (e.g., a correlation engine) based on the sensor accuracies and repeatability data mentioned above. Since the respective plausibility interval is not always constant but depends on the operating conditions, at least one suitable variable representing these conditions must be selected. This support variable is preferably also a robust measured variable that is approximately proportional to the engine's power output (e.g., fuel mass flow).

[0015] Preferably, when using information linking equations and / or mathematical calculation models, the presumably true value of a measured quantity is set as implausible if the presumably true value of this measured quantity lies outside an accuracy interval of a comparison value obtained with the associated information linking equation and / or outside an accuracy interval of a comparison value obtained with the associated mathematical calculation model.

[0016] It is particularly advantageous that, for the plausibility of the measured values ​​of at least some parameters of the test system, the recorded measurements are compared with the results of at least one information linking equation and at least one mathematical model for that parameter. Preferably, the presumed true value of a parameter is only considered implausible if it lies outside the accuracy interval of both a reference value obtained with the corresponding information linking equation and a reference value obtained with the corresponding mathematical model. Otherwise, the probably true value of the parameter in question is considered plausible.

[0017] In a further advantageous embodiment, it can be provided that, for the purpose of verifying the plausibility of the measured values ​​of at least some measured variables of the test system, the recorded measured values ​​are compared with the results of several information linking equations and / or several mathematical calculation models for these measured variables.

[0018] Preferably, the presumed true value of a measured quantity is considered plausible if it lies within an accuracy interval of at least one of the reference values ​​obtained using the associated information-linking equations, or within an accuracy interval of at least one of the reference values ​​obtained using the associated mathematical models. It is therefore not strictly necessary (but still possible) that the presumed true value lies within all accuracy intervals of the reference values ​​obtained using the associated information-linking equations and / or within all accuracy intervals of the reference values ​​obtained using the associated mathematical models.

[0019] Important prerequisites for plausibility checks using a logic tree, as used in the present invention, include, for example: • Basic tests report that the measuring and testing equipment has no obvious defects, • Standstill plausibility checks (for example, pressure=1 bar, speed and torque=0), • Setpoint / actual value comparisons (function conditioning), • Check whether the sensor devices regularly deliver measured values, • Trial calculations of the information linking equations and mathematical calculation models (test data set with comparison results), • Reference points / repetitions • Limit value monitoring for high operational reliability, • Gradient monitoring (thresholds for trends), • Outlier detection with exceedance of statistical parameters, • Detection of measurement range exceedances.

[0020] The presumed true value of a measured quantity—as it was measured—is the value provided by the "best" sensor. On its own, the presumed true value is unverifiable. Therefore, this value a) either to compare with a reference / comparison signal and / or b) to use for the plausibility check of measured values ​​of other measured quantities.

[0021] The order in which relationships between individual measurements are used in the plausibility checks follows the delimitability (i.e., which "category of relationship" can be used).

[0022] When establishing the chosen dependencies between the individual measured variables, the reliability of the method used (limitability and significance of the signal relationships) and the reliability of the quality of the measurement signals are important.

[0023] When mathematically representing relationships within mathematical models, models that depict the direction of action are preferred (i.e., the cause as input signals and the effect as output signals). Circular references should be avoided in the comparison signal generation and in the plausibility check logic.

[0024] Furthermore, it is important that multiple errors do not cancel each other out. The statement that a measured value is implausible, and vice versa, must not be made without justification.

[0025] Further features and advantages of the present invention will become clear from the following description of a preferred embodiment with reference to Fig. 1.

[0026] Fig. Figure 1 shows a logic tree which is used to verify the plausibility of the measured values ​​in a method for analyzing and evaluating measured values ​​of a test system according to a preferred embodiment of the present invention.

[0027] The test system is an engine test bench with multiple sensors that can acquire measurements of several variables. Using a logic tree implemented in an evaluation unit, which can be integrated into an existing measurement and testing system via a data transmission interface, a plausibility check can be performed for each of the measured variables according to predefined rules, by generating and using a presumptively true value.

[0028] During a test run, sensor readings of the measured variables are continuously acquired and automatically evaluated by the evaluation unit. To verify the plausibility of the presumed true values ​​of the measured variables, the measured values ​​are compared with reference values ​​provided in various ways.

[0029] One obvious approach is to provide at least two redundant sensors for each measured quantity and compare the measured values ​​to obtain the presumably true value of the respective quantity. However, this approach is not very practical in real-world applications, as it would require at least two sensors for every single measured quantity.

[0030] In the approach described here, only the measured values ​​of such parameters are compared using two or more redundant sensors, provided these parameters are expected to yield reliable, accurate, and reproducible results. A key concept is therefore to select specific parameters based on their reliability, taking into account measurement inaccuracies and the influence of the device under test, in order to define the degree of selectivity and the scope of a logical conclusion. Suitable parameters for this purpose include, for example, the rotational speed n of an internal combustion engine and its fuel consumption (mass flow rate) Bh.

[0031] The rotational speed n of the internal combustion engine is continuously measured using a primary speed sensor. A redundant secondary speed sensor S1 also acquires a comparative measurement of the rotational speed n. To determine the presumed true value of the rotational speed n, the measured values ​​of the primary speed sensor are compared with the comparative measurement of the secondary speed sensor S1. If the presumed true value of the rotational speed n, which is set based on the measurement acquired by the primary speed sensor, lies within an accuracy interval of the comparative value measured by the secondary speed sensor S1, then this presumed true value of the rotational speed n is plausible and very likely to be true.If the presumably true value of the rotational speed n lies outside the accuracy interval of the reference value measured using the second rotational speed sensor means S1, this presumably true value of the rotational speed n is not plausible.

[0032] The fuel consumption (mass flow rate) Bh of the internal combustion engine is continuously measured using a primary fuel consumption sensor. A redundant secondary fuel consumption sensor S2 also measures a comparative value of the fuel consumption Bh. To determine the presumed true value of the fuel consumption Bh, the measured values ​​of the primary fuel consumption sensor are compared with the comparative values ​​of the secondary fuel consumption sensor S2. If the presumed true value of the fuel consumption Bh, which is set based on the measured value obtained by the primary fuel consumption sensor, lies within an accuracy interval of the comparative value measured by the secondary fuel consumption sensor S2, then this presumed true value of the fuel consumption Bh is plausible and very likely to be true.If the presumed true value of fuel consumption Bh lies outside this accuracy interval, then this presumed true value of fuel consumption is not plausible.

[0033] A further sensor is used to acquire measurements of the air-fuel ratio λ of the internal combustion engine. To validate the measured values ​​of the air-fuel ratio λ, the Brettschneider equation B (see: BOSCH Technical Reports, Volume 6 (1979), serial no. 50277) and a mathematical model M1 are employed. This model receives as input the validated, presumed true values ​​of the engine speed n and fuel consumption Bh of the internal combustion engine, as well as charge exchange control variables, such as a throttle valve control variable DK and a camshaft intake control variable NWE. If the presumed true value of the air-fuel ratio λ lies within the accuracy interval of the reference value determined by the Brettschneider equation B or within the accuracy interval of the reference value determined by the mathematical model M1, the presumed true value is considered plausible.If the supposedly true value of the air number λ lies outside the accuracy interval of the comparison value determined by the Brettschneider equation B and outside the accuracy interval of the comparison value determined by the mathematical calculation model M1, it is not plausible.

[0034] The O2, CO2, and CO exhaust gas concentrations are also measured using suitable sensors. The O2 exhaust gas concentration is then validated. 1) by a mixture equation G1, which receives as its input the plausible, presumably true value of the air-fuel ratio λ, and 2) by a mixture equation G2, which receives the CO2 exhaust gas concentration as its input variable.

[0035] If the presumed true value of the O2 exhaust gas concentration lies within an accuracy interval of the reference value determined by mixture equation G1 or within an accuracy interval of the reference value determined by mixture equation G2, the presumed true value is plausible. If it lies outside the two accuracy intervals, it is implausible.

[0036] A plausibility check of the CO2 exhaust gas concentration is performed. 1) by a mixture equation G3, which receives as its input the plausible, presumably true value of the air-fuel ratio λ, and 2) by a mixture equation G4, which receives the O2 exhaust gas concentration as its input variable.

[0037] If the presumed true value of the CO2 exhaust gas concentration lies within an accuracy interval of the reference value determined by mixture equation G3 or within an accuracy interval of the reference value determined by mixture equation G4, the presumed true value is plausible. If it lies outside the two accuracy intervals, it is implausible.

[0038] A plausibility check of the CO exhaust gas concentration is performed. 1) by a mixture equation G5, which receives as its input the plausible, presumably true value of the air-fuel ratio λ, and 2) by a mixture equation G6, which receives the CO2 exhaust gas concentration as its input variable.

[0039] If the presumed true value of the CO exhaust gas concentration lies within an accuracy interval of the reference value determined by mixture equation G5 or within an accuracy interval of the reference value determined by mixture equation G6, the presumed true value is plausible. If it lies outside these two accuracy intervals, it is not plausible.

[0040] For determining the equations governing the mixture relationships, a direct, one-dimensional dependence on the exhaust gas concentrations is preferably established: •O2= f(Lambda) •CO2= f(Lambda) •CO2= f(O2) •CO = f(Lambda)

[0041] All equations are also used inversely. This relationship is determined based on selective variation measurements. Its representation is preferably not given as an absolute value equation, but relative to a basic equation. The basic equation represents the aforementioned functional relationship, which arises even without influencing engine combustion, but due to the dilution effects of the mixture. It is calculated based on a lambda-1 exhaust gas mixture, which is diluted towards "lean" by pure air and towards "rich" by pure fuel gas. Examples of its numerical values ​​in a polynomial equation can be found in the following representation:

[0042] Furthermore, pressure sensors are provided by means of which pressures in the combustion chamber of the internal combustion engine (for example, peak pressures or indicated mean pressures) as well as indicated pressures in the exhaust gas and indicated pressures in an intake system of the internal combustion engine can be measured. The measured values ​​for each of these pressures can be validated using a separate mathematical model M2. For the sake of simplicity, in Fig. Figure 1 presents a single mathematical model, M2, for validating pressure measurements. This mathematical model operates as an average value model over several operating cycles of the internal combustion engine. It receives input variables, state variables, and manipulated variables, and outputs an indicator variable. In this case, model M2 receives validated values ​​for the engine speed n and fuel consumption Bh, thus defining an operating point of the internal combustion engine. Furthermore, model M2 receives validated values ​​for the air-fuel ratio λ, ignition timing ZZP, valve timing manipulated variables, and optionally, temperature values. The output variable of model M2 is an indicator variable, such as the indicated mean effective pressure pmi.If the presumed true value of the indicated mean pressure (pmi) lies within an accuracy interval of the reference value determined by the mathematical model M2, then this presumed true value is plausible. Otherwise, it is not plausible.

[0043] A further sensor measures the torque of the combustion engine. Plausibility is checked using a mathematical model M3, which receives the validated values ​​of the rotational speed n and the indicated mean effective pressure pmi as inputs. If the presumed true value of the torque lies within the accuracy interval of the reference value determined by the mathematical model M3, this presumed true value is considered plausible. Otherwise, it is not.

[0044] Further sensors are provided to measure, for example, temperatures (media, air, exhaust gas), pressures (charge exchange system), HC exhaust gas concentrations, NOx exhaust gas concentrations, particle number concentrations, particle mass concentrations, and blow-by volume flows. To validate these measurements, a further mathematical model M4 is provided for each of these parameters, which is also an average model over several combustion engine cycles. For simplification purposes, in Fig.Figure 1 again shows only a single mathematical calculation model, M4, graphically represented. Model M4 receives input variables such as imprinted quantities, state variables, and indexed quantities. In this case, model M4 receives plausible values ​​for the rotational speed and fuel consumption Bh. These quantities define the operating point of the internal combustion engine. Furthermore, model M4 receives plausible values ​​for the air-fuel ratio λ and values ​​for an internal efficiency η. iand plausible values ​​of a compression end pressure, and optionally also temperature values. If the presumed true value of the measured quantity to be plausible lies within the accuracy interval of the reference value determined by the mathematical calculation model M4, this presumed true value is plausible. Otherwise, it is not plausible. Using the various model variants of model M4, blow-by volume flows, intake system temperatures, exhaust gas temperatures, exhaust system pressures, intake system pressures, HC, CO, and NO concentrations, as well as particle concentrations, can be plausibly verified. Optionally, it is also possible to re-plausibly verify previously monitored and otherwise plausible quantities, such as torque and O2 and CO2 exhaust gas concentrations.

[0045] It becomes clear that each of the mathematical calculation models M1-M4 receives a number of input variables. Some of these input variables were validated during the execution of the logic tree by determining the likely true value for these measured variables. Some of the input variables are actuator variables, such as a throttle position (TPV), a camshaft intake position (CAP), or the ignition timing (IQP).

[0046] For the mathematical models M1-M4, it is important to note that the accuracy interval is determined by the accuracy of the input signals (measured values) in relation to the output signals (input variance), as well as by inaccuracies in the modeling itself (so-called residuals). The uncertainty of the repeatability of a result is considered superimposed, necessitating a case-by-case distinction (depending on which uncertainty is greater) to determine which interval is used. In each case, the larger of these intervals is used as the determining factor.

[0047] Plausibility checks using redundant sensors cannot detect test object defects, as they measure and validate signals that are not attributable to faulty measurement technology. However, errors can be detected by validating the measured values ​​of specific parameters using information-linking equations or mathematical models. Initially, it is not possible to determine whether the fault lies with the test object or the measuring and testing equipment. Nevertheless, the number of errors provides useful clues about the cause. For example, if only a single error occurs, it is highly likely that only the corresponding sensor is malfunctioning. If numerous errors are diagnosed, this suggests that the test object itself is the source of these errors.Because it is rather unlikely that several sensor devices would be defective at the same time.

[0048] The method presented here enables plausibility assessments of the measured values ​​of individual parameters to be made during a test run. This allows for a determination of the plausibility of the measured values ​​for all considered parameters even during the test run ("online"). The results remain available ("offline") even after the test run has ended, allowing for subsequent plausibility checks based on stored data.

[0049] The procedure described here is based on the following questions and considerations. 1) “Viewpoint” from the direction of the signals in question: From which other measurement signal can a statement be made regarding the measurement signal in question, or what signal relationships exist? With which methodological means can a statement be made ("limitation")? The use of the signals according to their "reliability," taking into account measurement inaccuracies and influences of the test specimen, is also essential for determining the discriminatory power and scope of a statement. 2) Introduction of a logic tree using the most direct relationships possible: Signal processing is used as little as possible. Mixture equations are preferred over balance equations. Passive observer systems are used for simplification. Process models are only included where absolutely necessary. 3) Formation and use of the hypothesis of the presumably true value of a measured quantity. 4) The quality statement itself is used for the persistence statement (time interval as query limit). The diagnoses are always active. A release component is not provided.

Claims

[1] Method for analyzing and evaluating measured values ​​of a test system designed as an engine test bench in which internal combustion engines are examined, wherein a test specimen is monitored during a test run by means of a plurality of sensor means and each of the sensor means provides measured values ​​of at least one measured quantity to an evaluation unit for further processing, characterized by , that a logic tree is generated to verify the plausibility of the measured values ​​of the measured quantities, by means of which a plausibility is determined for each of the measured quantities according to predefined rules by forming and using a presumably true value. [2] Method according to claim 1, characterized by , that redundant measurements are carried out using a number of redundant sensor devices to verify the plausibility of at least some measured variables of the test system. [3] Method according to claim 2, characterized by, that the presumed true value of a measured quantity is set as implausible if the presumed true value of this measured quantity lies outside an accuracy interval of a reference value measured with at least one redundant sensor means. [4] Method according to any one of claims 1 to 3, characterized by , that in order to verify the plausibility of the measured values ​​of at least some measured variables of the test system, the recorded measured values ​​are compared with the results of at least one information linking equation for this measured variable. [5] Method according to any one of claims 1 to 4, characterized by , that in order to verify the plausibility of the measured values ​​of at least some of the parameters of the test system, the recorded measured values ​​are compared with the results of at least one mathematical calculation model for these parameters. [6] Method according to one of claims 4 or 5, characterized by, that the presumably true value of a measured quantity is set as implausible if the presumably true value of this measured quantity lies outside an accuracy interval of a comparison value obtained with the associated information linking equation and / or outside an accuracy interval of a comparison value obtained with the associated mathematical calculation model. [7] Method according to one of claims 4 or 5, characterized by , that in order to verify the plausibility of the measured values ​​of at least some of the measurement variables of the test system, the recorded measured values ​​are compared with the results of at least one information linking equation and at least one mathematical calculation model for this measurement variable. [8] Method according to claim 7, characterized by, that the presumably true value of a measured quantity is only considered implausible if the presumably true value of this measured quantity lies both outside an accuracy interval of a comparison value obtained with the associated information linking equation, and outside an accuracy interval of a comparison value obtained with the associated mathematical calculation model. [9] Method according to one of claims 4 or 5, characterized by , that in order to verify the plausibility of the measured values ​​of at least some measured variables of the test system, the recorded measured values ​​are compared with the results of several information linking equations and / or several mathematical calculation models for this measured variable. [10] Method according to claim 9, characterized by, that the presumably true value of a measured quantity is considered plausible if the presumably true value lies within an accuracy interval of at least one of the comparison values ​​obtained with the associated information linking equations or within an accuracy interval of at least one of the comparison values ​​obtained with the associated mathematical calculation models.

Images