Method and device for testing a component part of an aircraft

Abstract

A method for testing a component part of an aircraft comprises the steps of determining at least one first test value of the component part of the aircraft and/or at least one first test value of a comparable component part of a further aircraft for at least one test parameter, inputting the first test parameter and the first test value into a pattern recognition system, which produces an inner correlation between the first test parameter and the first test value. The method further comprises the steps of defining at least one second test parameter, inputting the second test parameter into the pattern recognition system in order to determine a second test value by means of the inner correlation, checking whether the second test value falls within the predefined value range, and determining a third test value of the component part of the aircraft for the second test parameter if the second test value falls within the predefined value range. The invention further relates to a device for testing a component part of an aircraft.

Claims

1. A method for producing inner correlations between test parameters and test values and for selecting test parameters for testing a component part of an aircraft, said method comprising the steps of: a) measuring, using at least one measurement arrangement, a first test value for a first test parameter of the component part of the aircraft, said first test parameter representing a specification for changing or manipulating the component part and said measuring first test value is performed during a change or manipulation of said component part corresponding to the first test parameter; b) automatically inputting said first test parameter and said first test value into a pattern recognition system, and producing an inner correlation between said first test parameter and said first test value, said inner correlation representing a relationship between said first test parameter and said first test value; c) defining at least one second test parameter of the component part of the aircraft using a test parameter definition procedure; d) automatically inputting said second test parameter into said pattern recognition system in order to determine a second test value, wherein the second test value is a predicted value based on the inner correlation; e) checking whether the second test value falls within a predefined value range; and f) measuring, using the at least one measurement arrangement, a third test value of said component part of the aircraft for the second test parameter when the second test value falls within the predefined value range.

2. The method as claimed in claim 1, wherein the first test value and/or the third test value is/are measured by a mechanical and/or electrical measurement, and in order to determine the third test value, the first test parameter further being replaced by the second test parameter if the second test value falls within the predefined value range, the second test value being determined from the second test parameter which replaces the first test parameter, or from the first test parameter and the second test parameter which replaces the first test parameter.

3. The method as claimed in claim 1, wherein the first test parameter and/or the second test parameter comprise/comprises at least one component from the following group: extent of manipulation of the component part, location of manipulation of the component part, method of manipulation of the component part or duration of manipulation of the component part.

4. The method as claimed in claim 1, wherein the pattern recognition system is formed as a neural network.

5. The method as claimed in claim 1, wherein, in step c), the test parameter definition procedure defines the second test parameter randomly or in a manner based on a stochastic distribution.

6. The method as claimed in claim 1, wherein the test parameter definition procedure includes before step c), forming the center point of a cluster of first test parameters or of first test values, the center point being used as a starting point for the definition of the second test parameter, or using the first test values which fall within the predefined value range as starting points for the definition of the second test parameter.

7. The method as claimed in claim 1, wherein step c) is provided by a genetic algorithm, the second test parameter being generated from the combination of two or more first test parameters, the type and manner of the combination further being performed randomly.

8. The method as claimed in claim 1, wherein the predefined value range is a range of a probability distribution which specifies the probability for a fault of the component part according to the second test parameter, the probability for a fault for the predefined value range being greater than 80%, or being less than 20%.

9. The method as claimed in claim 1, wherein the component part is a control arrangement, in particular a control arrangement for an ambient control system of a cabin of the aircraft or a control arrangement for steering the aircraft, the first test parameter and/or the second test parameter comprising: external temperature, internal temperature in the cabin, altitude of the aircraft, aircraft weight and/or center of gravity of the aircraft.

10. The method as claimed in claim 1, wherein the component part is a power supply arrangement, which provides power at various supply points in the cabin and/or at a kitchen of the aircraft, the first test parameter and/or the second test parameter comprising: the moment of connection of the power consumption at a supply point, the moment of disconnection of the power consumption at a supply point, the power consumption at a supply point, the number of interruptions of the power supply between two supply points, and/or the duration of the interruptions of the power supply between two supply points.

11. The method as claimed in claim 1, wherein the component part is an electrically operated valve, the first test parameter and/or the second test parameter comprising the moment of interruption of the power supply of the valve, the number of power interruptions, the duration of the power interruptions, and/or the interval between two power interruptions.

12. The method according to claim 1, further comprises automatically inputting the determined second test parameter and the third test value into the pattern recognition system to replace the first test parameter and the first test value respectively and produce another inner correlation between the first test parameter and first test value.

13. The method according to claim 1, wherein the steps a) to f) are performed in a specified order.

14. The method according to claim 1, wherein the steps a) to f) are carried out repeatedly to train the produced inner correlation and select further second test parameters.

15. The method according to claim 1, wherein the steps a) to b) are carried out repeatedly to train the produced inner correlation and select further second test parameters.

16. A device for producing inner correlations between test parameters and test values and for selecting test parameters for testing a component part of an aircraft, comprising a measuring arrangement, which is suitable for measuring a first test value for a first parameter, the first test parameter representing a specification for changing or manipulating the component part and the measuring arrangement measuring the first test value during a change or manipulation of the component part corresponding to the first test parameter, a pattern recognition system, which is suitable for producing an inner correlation between the first test value and the first test parameter, the inner correlation representing a relationship between the first test parameter and the first test value, and for generating a second test value from a second test parameter by predicting the second test value based on the inner correlation, a test parameter definition arrangement, which defines the second test parameter of the component part of the aircraft, a selection arrangement, which is suitable for dividing a second test value into classes, the measuring arrangement measuring a third test value on the basis of the second test parameter if the second test value falls within a predefined one of the classes.

17. The device as claimed in claim 16, wherein the measuring arrangement determines the first test value and/or the third test value by a mechanical and/or electrical measurement.

18. The device as claimed in claim 16, wherein a plurality of first test parameters are provided, the test parameter definition arrangement defining a plurality of second test parameters, and the selection arrangement further replacing at least one first test parameter by at least one second test parameter if the second test value falls within the predefined class, the pattern recognition system determining the second test value from the second test parameter which replaces the first test parameter, or the pattern recognition system determining the second test value from at least one first test parameter and the at least one second test parameter which replaces the first test parameter.

19. A method for producing correlations between test parameters and test values comprising: measuring a first test value of the component part of an aircraft wherein the first test value corresponds to a first test parameter representing a specification for changing or manipulating the component part, wherein the first test value is measured during a change or manipulation of the component part corresponding to the first test parameter; automatically inputting the first test parameter and the first test value into a pattern recognition system, and producing an inner correlation between the first test parameter and the first test value; defining a second test parameter of the component part of the aircraft using a test parameter definition procedure; automatically inputting the second test parameter into the pattern recognition system to produce a second test value, wherein the second test value is a predicted value based on the inner correlation; determining whether the second test value falls within a predefined value range; and in response to the determination that the second test value falls within the predefined value range, measuring a third test value of the component part of the aircraft, wherein the third test value corresponds to the second test parameter.

20. The method of claim 19, further comprises automatically inputting the determined second test parameter and the measured third test value into the pattern recognition system to replace the first test parameter and the first test value respectively and produce another inner correlation between the first test parameter and first test value.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) The invention will be characterized in greater detail hereinafter on the basis of exemplary embodiments. In the schematic drawings

(2) FIG. 1 shows a block diagram which defines an embodiment of the method;

(3) FIG. 2 shows a schematic illustration of a device for testing a component part of an aircraft;

(4) FIG. 3 shows a graph which presents test values according to test parameters;

(5) FIG. 4 shows a schematic illustration of the aircraft with an electrical power system;

(6) FIG. 5 shows a schematic illustration of the aircraft with an ambient control system;

(7) FIG. 6 shows a schematic illustration of the aircraft with a valve; and

(8) FIG. 7 shows a schematic illustration of the aircraft with a flight control system.

DETAILED DESCRIPTION OF EMBODIMENTS

(9) The block S1 in FIG. 1 stands for a step in which a component part 10 of an aircraft 40 is tested. For this test, first test parameters are used which deliver at least one first test value. The step S1 corresponds to step a). The at least one first test parameter and the at least one first test value are then analyzed in step S2. This occurs here by clustering of the test values. In other words, test values are grouped to form clusters. Here, the clustering is implemented in accordance with test results, such as fault messages or warning messages. A new starting point for a new, second test parameter is determined on the basis of the centerpoint of the cluster of the test values. The starting test value thus obtained is recalculated into test parameters. In step S3, which corresponds to step c) in claim 1, the second test parameter is defined by this starting point. This occurs randomly. The chance determines the distance of the second test parameter from the starting point test parameter. Alternatively, the second test parameter can be established by fixed steps with predefined step increments proceeding from the starting point.

(10) After step S1, step S4 is also performed simultaneously with steps S2 and S3. In step S4, which corresponds to step b) in claim 1, the first test parameter and the first test value are input into a pattern recognition system. The pattern recognition system then produces an inner correlation between the first test parameter and the first test value. This step can also be referred to as a learning step.

(11) In step S5 this inner correlation is used to determine a second test value on the basis of the second test parameter. This corresponds to step e) in claim 1. In step S6 all those second test parameters from the plurality of second test parameters generated in step S3 of which the second test value established in step S5 falls within the predefined value range are then selected. This corresponds to step e) in claim 1. These second test parameters thus determined are then used in a new test method in step 1 as test parameters in order to determine the third test value.

(12) A device for testing a component part 10 of an aircraft 40 comprises a measuring arrangement 12, a pattern recognition system 14, a test parameter definition arrangement 16 and a selection arrangement 18.

(13) The measuring arrangement 12 has a measuring head 20, by means of which the measuring arrangement 12 can take measurements on the component part 10. The measuring arrangement 12 further has a manipulation arrangement 22, by means of which the measuring arrangement 12 can change the component part 10. The change is implemented in order to carry out a test with respect to the functional capacity of the component part 10. Further, the manipulation arrangement 22 may also change external factors which act on the component part 10, for example pressure, heat or applied voltage. The effect of changes to the component part 10 can then be determined by means of the measuring head 20. The test parameters are therefore used by the manipulation arrangement 22. The first test values are output by the measuring head 20 to the pattern recognition system 14 and to the test parameter definition arrangement 16.

(14) The test parameter definition arrangement 16 comprises an analysis arrangement 24 and a random device 26. The analysis arrangement 24 determines the starting point, on the basis of which the second test parameter is defined by the random device 26. In the exemplary embodiment presented here, the analysis arrangement 24 forms the centerpoint of a cluster of first test values. The centerpoint is converted into the corresponding test parameters. The random arrangement 26, by means of a random generator, determines the distance from the test parameter that has been determined by the analysis arrangement 24. The starting point plus the distance is then the second test parameter.

(15) The first test parameter and the first test value are input as inputs into the pattern recognition system 14. The pattern recognition system 14 determines the inner correlation between the first test value and the first test parameter. The inner correlation is then used in order to determine the second test value from the second test parameter which is input by the test parameter definition arrangement 16 into the pattern recognition system 14.

(16) The second test value is then fed to the selection arrangement 18. This divides the second test value into a number of classes. If the second test value from the selection arrangement 18 falls within the predefined class as determined previously, the first test parameter is replaced by the second test parameter. The second test parameter is again fed to the measuring arrangement 12, which determines the third test value.

(17) A practical example, which is shown in FIG. 3, is a test method which was developed in order to simulate a component part 10 for fault simulation and to test said component part. The simulation is representative in terms of the test results. The test was a power supply interruption test, which is directed at testing the robustness of a system to be tested by implementing a number of power interruptions. The test parameters which characterize the power interruption are as follows: number of interruptions, duration of an individual interruption, and duration for which the system to be tested was connected before the first interruption occurred. More specifically, the test parameters (first test parameters) are as follows: number of interruptions (test parameter definition) [1, 10], step size 1: (1; 2; 3; 4; 5; . . . 10;) (test parameter size) interruption duration (test parameter definition) [1, 20], step size 5: (1; 6; 11; 16; 20;) (test parameter size) time until first interruption (test parameter definition) [1, 20], step size 5: (1; 6; 11; 16; 20;) (test parameter size)

(18) There are thus a total of 4,000 points or individual test parameter sets.

(19) The test values are as follows:

(20) TABLE-US-00001 0 normal point 1 warning point, that is to say probable candidate for faults 2 fault point.

(21) A function test has the following results:

(22) There are four fault zones 30 with: 6 faults 96 warnings

(23) The fault zones 30 are distributed in the space of the test values which are produced from the test parameters. The fault zones 30 are grouped around faults 34, around which warnings 32 are arranged. The warnings 32 and the fault points 34 are examples for first test values. The warnings 32 and the faults 34 are those first test values that fall within the predefined range. The predefined range in this exemplary embodiment therefore consists of faults and warnings.

(24) The pattern recognition system can recognize the correlation between the first test parameter (number of interruptions . . . ) and the first test values. The centerpoint of a fault zone 30 can be used as a starting point. The new second test parameter can be defined by a random displacement from the starting point. The new second test parameter is input by the pattern recognition system in order to determine whether the second test value also lies within the fault zone 30. If the second test value 30 lies in the fault zone 30, the second test parameter corresponding to this second test value is used for the new test method.

(25) Some exemplary embodiments will be presented briefly hereinafter.

(26) Cabin Electrical Power System

(27) The electrical power system 42 of an aircraft 40 shown in FIG. 4 is responsible for providing power at the passenger seats 44 and in the aircraft kitchen. The electrical power system 42 is an example of a power supply arrangement. The large number of different power tap points 46 and also the different load and the different load distribution per tap point result in a large state space in which testing must be performed. The power tap point 46 is an example of a supply point. The power is generated by a generator 48 and is fed via a cable 50 to the power tap points 46.

(28) Test-Relevant Input Parameters:

(29) moment of connection and disconnection of the consumers at the individual power tap points 46 (sockets)

(30) load per power tap point 46 at moment in time t

(31) Component test (valve, power interrupt test)

(32) With an electrically operated and controlled valve 52, which is shown in FIG. 6, the behavior can be influenced by interruptions of the power supply during the take-off process. The valve 52 for example may control the airflow between a cabin 61 and the surroundings of the aircraft 40. The power supply is provided by a generator 48, which supplies the valve 52 with energy via the cable 50.

(33) Test-Relevant Input Parameters:

(34) time until the first power interruption

(35) number of power interruptions

(36) duration of a power interruption

(37) time between the power interruptions

(38) Air Conditioning System

(39) An air conditioning system, shown in FIG. 5, in aircraft 40 (environmental control system or ECS 54) comprises the following three system components: air exchange 56, pressure control 58 and temperature control 60 in the cabin 61 of the aircraft 40 for crew, passengers and luggage compartments. The air conditioning system is required in aircraft 40 in order to provide the necessary atmosphere in the cabin 61 at altitudes up to more than 11,000 meters and to provide the passengers with sufficient air pressure, a sufficient oxygen supply and an appropriate ambient temperature.

(40) Compared with normal air conditioning systems only for temperature control, for example in buildings or vehicles, the use of this term for the ECS in aircraft is incomplete, because pressure and oxygen are also supplied here. The air conditioning systems in aircraft 40 therefore differ from the conventional air conditioning systems by a different design and energy source with much greater power demand and high safety requirements.

(41) Test-Relevant Input Parameters:

(42) number of passenger seats 44

(43) degree of opening of the nozzle per passenger seat 44

(44) external temperature

(45) current internal temperature

(46) target internal temperature

(47) altitude

(48) target cabin height (internal pressure)

(49) (Fly-by-Wire) Flight Control System

(50) The flight control system shown in FIG. 7 is an example of a control arrangement for steering the aircraft 40. The inputs of the pilot at the control elements (for example at the sidestick 62) are converted in the case of fly-by-wire by means of a convertor 64 into electrical signals, which are then in turn converted into movements of the control surfaces 68 by servomotors 66 or by hydraulic cylinders, which are controlled by means of electric valves. This is necessary since, in the case of large or quick aircraft 40, the force to be applied by the pilot in order to move the control surfaces would be unreasonably large, or complex power transmissions would be uneconomical.

(51) Test-Relevant Input Parameters:

(52) (auto)pilot input (thrust, steering command, . . . )

(53) angle of attack (flap setting)

(54) aircraft weight

(55) altitude

(56) external temperature

(57) current center of gravity

(58) Aspects of the invention will lastly be presented by way of example:

(59) Methods from the field of machine learning/artificial intelligence, specifically neural networks and/or genetic algorithms, are linked with methods from the field of data mining/knowledge discovery from databases, specifically clustering, and existing methods for test parameter generation, specifically random data generation, for the adaptive testing.

(60) On the condition that critical test parameters are usually to be found in groups in the entire test parameter space, optimized test parameters for the next test run can thus be generated starting from tests already performed. The starting point for the adaptive testing is always the definition of all possible stimuli for the tested system (system under test, SUT).

(61) The uniqueness of the adaptive testing then becomes significant if first test values are already present. Should this not be the case, the adaptive testing then corresponds largely to the test method used for the generation of the new test parameters, that is to say without test values and with use of random generators for the generation of new test parameters, the adaptive testing corresponds to the random-based testing.

(62) 1. Analysis of the Test Data

(63) The objective is to obtain starting points for the generation of new test parameter values. Various methods can be used for this purpose:

(64) clustering: test values are grouped into clusters. The new starting point may then be the centerpoint of a cluster, for example.

(65) fault points: all found fault points are a starting point

(66) 2. Generation of New Test Parameters

(67) New potential new test parameters are generated on the basis of the results of the test value analysis, starting from the previously established starting points. Various methods can be used for this purpose, for example:

(68) random: new values are generated in a purely randomized manner stochastic: new values are generated based on stochastic distributions (for example Gaussian distribution) step increments: new values are obtained by steps having predefined step increments

(69) 3. Learning and Prognosis

(70) Result data of tests that have already been carried out are used by a neural network in order to learn a decision function, which can perform a classification in output data automatically on the basis of the input data. For example, input data can be classified in output data classes, such as no fault, warning or fault. The learnt decision function can then be used to automatically classify new input data. New test data are data that have not yet been used previously for learning. The accuracy of the classification is dependent on the structure of the neural network (for example mono-layer or multi-layer network) and on the type of activation function (for example linear or non-linear). A good and useful approximation generally aspires for new test data to be classified with a low fault on the basis of training data already learnt. Put more simply, the trained neural network is used to make a prediction, for new test parameter values, with regard to which value combinations will likely lead to faults. This information is used in the next step to generate new suitable test parameters.

(71) 4. Selection of the Test Parameters

(72) When generating new test parameters, a greater number of combinations are usually generated, in particular if a number of parameters are combined with the Cartesian product. Since it is not always possible to carry out thousands of tests, the final number of new test parameters can be reduced in this step. This can be achieved in accordance with different viewpoints: only parameters with high fault probability mixture of parameters with high and low fault probability absolute upper limits of the test parameter number

(73) Experiments have shown that, by the adaptive testing, fewer tests are sufficient in order to find just as many or more faults in a component as by the previously applied methods. Furthermore, the test parameter generation is nowadays also often carried out manually, whereas the adaptive testing is directed to automatic testing. After each test run with a set of test parameters, the resultant test values are used to adjust/to adapt the test focus for the next test run.

LIST OF REFERENCE SIGNS

(74) 10 component part 12 measuring arrangement 14 pattern recognition system 16 test parameter definition arrangement 18 selection arrangement 20 measuring head 22 manipulation arrangement 24 analysis arrangement 26 random device 30 fault zone 32 warning 34 fault 40 aircraft 42 electrical power system 44 passenger seat 46 power tap point 48 generator 50 cable 52 valve 54 ambient control system 56 air exchange 58 pressure control 60 temperature control 61 cabin 62 sidestick 64 converter 66 servomotor 68 control surface S1 step 1 S2 step 2 S3 step 3 S4 step 4 S5 step 5 S6 step 6