CN116644650A - Computer-implemented method for evaluating structure-borne noise - Google Patents
Computer-implemented method for evaluating structure-borne noise Download PDFInfo
- Publication number
- CN116644650A CN116644650A CN202310055110.7A CN202310055110A CN116644650A CN 116644650 A CN116644650 A CN 116644650A CN 202310055110 A CN202310055110 A CN 202310055110A CN 116644650 A CN116644650 A CN 116644650A
- Authority
- CN
- China
- Prior art keywords
- training
- dataset
- evaluation
- artificial intelligence
- data set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 20
- 238000012549 training Methods 0.000 claims abstract description 63
- 238000011156 evaluation Methods 0.000 claims abstract description 51
- 238000013473 artificial intelligence Methods 0.000 claims abstract description 48
- 230000001419 dependent effect Effects 0.000 claims abstract description 3
- 230000002950 deficient Effects 0.000 claims description 20
- 238000012545 processing Methods 0.000 claims description 7
- 238000004519 manufacturing process Methods 0.000 description 8
- 230000007547 defect Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 2
- 230000000712 assembly Effects 0.000 description 1
- 238000000429 assembly Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M7/00—Vibration-testing of structures; Shock-testing of structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/15—Vehicle, aircraft or watercraft design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/10—Noise analysis or noise optimisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
Abstract
The invention relates to a computer-implemented method for evaluating structure-borne noise, comprising the steps of: -recording a plurality of training data sets (S3) of structure-borne noise data, wherein the training data sets are each recorded when a force acts on a component, and the structure-borne noise data is dependent on the force; -training artificial intelligence using the training dataset (S5); -recording a plurality of first evaluation data sets of structure-borne noise data (S7); -evaluating the first evaluation dataset using the trained artificial intelligence; and subsequently-recording at least one first calibration data set of structure-borne noise data; -training the artificial intelligence using the at least one first calibration data set (S8).
Description
Technical Field
The present invention relates to a computer-implemented method for evaluating structure-borne noise.
Background
In the manufacture of motor vehicles, structure-borne noise data recorded by structure-borne noise sensors are evaluated in order to detect defective components and/or defective component assemblies. A method is known from DE 10 2019 003 679A1 in which a machine learning algorithm is used to evaluate engine anomalies.
Disclosure of Invention
In contrast, the problem addressed by the present invention is to create a method that adapts to changing environments. In addition, a system will be created that can perform such a method.
This problem is solved by a method and system according to the following embodiments.
In a method according to one embodiment, a plurality of training data sets of structure-borne noise data are recorded. This may be done, for example, with a sensor. The training data sets are each recorded when a force is applied to the component. The noise data is dependent on the force. For example, each training data set may correspond to a component and a varying force.
Artificial intelligence is trained using a training data set. This may be supervised, unsupervised or partially supervised, for example. In particular, in the context of the present specification, supervised training is understood to mean that certain parameters are assigned to a training dataset, for example, prior to training. This may be done, for example, by the user. For example, the parameter may be that the training data set in question belongs to a non-defective or defective component. In the context of the present specification, defective components are also understood to mean that the respective components are not assembled correctly, for example wrongly or inadequately.
In this way, artificial intelligence is trained to detect defective parts, regardless of the cause of the defect. The reason for this is that training of artificial intelligence is not intended to detect specific defects, but is generally intended to detect deviations between datasets belonging to non-defective parts and datasets belonging to defective parts. However, in another embodiment of the present invention, it is possible to train artificial intelligence to detect specific defect causes.
In addition, a plurality of first evaluation data sets of structure-borne noise data are recorded. This may be done, for example, using the same sensor as the recorded training dataset or using other sensors. The first evaluation data set may not be recorded under exactly the same conditions as the training data set. For example, the training data set may be recorded under conditions that have particularly little impact on the recorded structure-borne noise data. For example, the first evaluation dataset may be recorded on the production line or at the end of the production line. In this way, vibrations occurring in the production line may affect the recording of the first evaluation dataset.
The assessment dataset is assessed using trained artificial intelligence.
Thereafter, at least a first calibration data set of structure-borne noise data is recorded. This may be done, for example, using the same sensor as the recording of the first evaluation dataset or using other sensors. It is possible that at least one first calibration data set is recorded under conditions which are not exactly the same as the first evaluation data set and/or the training data set. For example, at least one first calibration data set may be recorded under conditions that have particularly little impact on the recorded structure-borne noise data.
The artificial intelligence is trained using at least one first calibration data set. This is done although artificial intelligence has been trained on the training data set. This is particularly advantageous because artificial intelligence can thus accommodate any unnoticed changes in conditions and/or components.
According to one embodiment of the invention, prior to training the artificial intelligence using the training data set, the artificial intelligence may classify the training data set into clusters based on the quality of the training data set. Only training data sets with quality above the limit are used for training the artificial intelligence.
According to an embodiment of the invention, after said training, a second evaluation dataset of structure-borne noise data may be recorded using said at least one first calibration dataset, and said second evaluation dataset is evaluated using said artificial intelligence. Thus, the second evaluation dataset may be evaluated using artificial intelligence that adapts to any changing environment and/or component.
According to one embodiment of the invention, in the evaluation, the first evaluation data set and/or the second evaluation data set is evaluated by the artificial intelligence as being associated with a non-defective part or with a defective part. Here, it is possible for artificial intelligence to evaluate the corresponding evaluation dataset as belonging to a defective component, regardless of the cause or type of defect. It is particularly advantageous that unknown disturbance variables are also detected as disturbances, since training is performed using data sets belonging to non-defective parts.
According to one embodiment of the invention, only structure-borne noise data from non-defective parts may be used in order to record the at least one first calibration data set.
According to one embodiment of the invention, the first type of component may be used in each case when recording the training data set, the first and second evaluation data sets and the at least one first calibration data set. In the context of the present specification, the type of component is specifically understood to mean a specific design. For example, in a single series of productions, multiple components of the first type may be manufactured in the same manner throughout. In particular, this may always be the same component of a single vehicle model.
After said evaluation of said second evaluation dataset, second calibration datasets are each recorded using a second type of component. In contrast, no second type of component is used when recording the training data set, the first and second evaluation data sets and the at least one first calibration data set. The artificial intelligence is trained using the second calibration data set.
This embodiment is particularly advantageous for adapting trained artificial intelligence to another purpose, namely for evaluating noise data of a component of the second type. This is very likely, especially when the first type of component has only a relatively small difference compared to the second type of component. For example, when a first type of component is used in a first motor vehicle model for a particular purpose and a second type of component is used in a second motor vehicle model for the same purpose, training using the second calibration data set is sufficient to evaluate the structure-borne noise data of the second type of component. Advantageously, entirely new training of the artificial intelligence or of another artificial intelligence may be omitted.
According to one embodiment of the invention, a second type of component is used to each record a third set of evaluation data of the structure-borne noise data. The third evaluation data set may be evaluated after training the artificial intelligence using the second calibration data set. For example, the evaluating may include the third evaluation dataset being evaluated as belonging to a defective or non-defective part.
According to one embodiment of the invention, the component may be a part of a motor vehicle. In particular, all the components mentioned in this description are possible components of a motor vehicle.
According to an embodiment of the invention, the movement of the respective component relative to the other component, the strength of the respective component and/or the presence of an imbalance may be evaluated when evaluating the first evaluation data set, the second evaluation data set and/or the third evaluation data set.
The system includes a digital electronic storage medium and a digital electronic processing unit. For example, the processing unit may be a digital signal processor. Instructions are stored in the storage medium. The processing unit is configured to facilitate reading and execution of the instructions. The processing unit is further designed to implement the method according to an embodiment of the invention when executing the instructions.
Drawings
Other features and advantages of the present invention will become apparent from the following description of preferred exemplary embodiments, which proceeds with reference to the accompanying drawings. The same reference numerals are used for the same or similar features and features having the same or similar functions. Here:
FIG. 1 shows a schematic block diagram of training artificial intelligence according to one embodiment of the invention; and is also provided with
FIG. 2 shows a schematic block diagram of evaluating structure-borne noise using trained artificial intelligence.
Detailed Description
In step S1, a sensor for recording structure-borne noise data is first arranged and/or fastened in place in order to record the structure-borne noise waves of the component to be tested. In an optional step S2, the sensor may be connected to a control unit. The control unit is configured to facilitate control of the recording of the training data set from the structure-borne noise data recorded by the sensor and possibly also to synchronize the recording of the training data set from the structure-borne noise data recorded by the sensor.
In step S3, the recording of the training data set is performed from the structure-borne noise data recorded by the sensor. In optional step S4, the training data set may be classified into clusters based on its quality. This may be accomplished by the artificial intelligence to be trained. In particular, this may be unsupervised learning.
In step S5, the artificial intelligence is trained using the training data set. When step S4 is performed, it is possible that only training data sets with quality higher than the limit value will be used for training. In step S6, a training model for use with the artificial intelligence is stored in a digital memory.
When artificial intelligence is used to evaluate the structure-borne noise data, a first set of evaluation data of the structure-borne noise data is recorded in step S7. This may be implemented using the structure used in step S3. However, it is also possible to use different sensors in order to record the first evaluation dataset. Specifically, step S7 may be performed within the production line or at the end of the production line, as compared to step S3. In this case, the recording of the first evaluation data set may be affected by vibrations and noise that are not present when recording the training data set. However, artificial intelligence may evaluate the first evaluation dataset. Specifically, the artificial intelligence evaluates whether the first evaluation dataset belongs to a non-defective part or a defective part.
After a certain number of evaluations of the first evaluation data sets, at step S8, artificial intelligence is trained using at least one first calibration data set. Specifically, it is possible to periodically perform step S8. By training using at least one first calibration data set, the artificial intelligence is adapted to potentially changing environmental conditions or components that change due to changing production processes.
It is also possible to train artificial intelligence using a plurality of second calibration data sets in step S8. In this case, for example, the second calibration data set may be associated with a different type of component than the training data set. By training using the second calibration data set, the artificial intelligence may be adapted to evaluate the evaluation data set in relation to other types of components. This is very likely, especially when the difference between the second calibration data set and the components of the training data set is relatively small. In this case, step S8 may replace the complete training of artificial intelligence to evaluate other types of components.
In step S9, if at least one first calibration data set is used in step S8, artificial intelligence may be used in order to evaluate the second evaluation data set. The second evaluation data sets are each associated with a component of the same type as the component corresponding to the first evaluation data set, respectively. If the second calibration data set is used in step S8, artificial intelligence may be used in step S9 in order to evaluate a third evaluation data set related to other types of components.
Claims (10)
1. A computer-implemented method for evaluating structure-borne noise, comprising the steps of:
-recording a plurality of training data sets (S3) of structure-borne noise data, wherein the training data sets are each recorded when a force acts on a component, and the structure-borne noise data is dependent on the force;
-training artificial intelligence using the training dataset (S5);
-recording a plurality of first evaluation data sets of structure-borne noise data (S7);
-evaluating the first evaluation dataset using the trained artificial intelligence; and then
-recording at least one first calibration data set of structure-borne noise data;
-training the artificial intelligence using the at least one first calibration data set (S8).
2. The method according to claim 1, wherein prior to training the artificial intelligence using the training data set, the artificial intelligence classifies the training data sets into clusters based on the quality of the training data set (S4), wherein only training data sets with quality higher than a limit are used for training the artificial intelligence.
3. The method according to any of the preceding claims, characterized in that after the training a second evaluation dataset of structure-borne noise data is recorded using the at least one first calibration dataset and the second evaluation dataset is evaluated using the artificial intelligence (S9).
4. Method according to any of the preceding claims, characterized in that in the evaluation the first evaluation data set and/or the second evaluation data set are evaluated by the artificial intelligence as being associated with a non-defective part or with a defective part.
5. Method according to any of the preceding claims, characterized in that only structure-borne noise data from non-defective parts is used in order to record the at least one first calibration data set.
6. The method according to any of the three preceding claims, characterized in that in recording the training dataset, the first and second evaluation dataset and the at least one first calibration dataset, a first type of component is used in each case, wherein after the evaluation of the second evaluation dataset, a second calibration dataset is each recorded using a second type of component, wherein in recording the training dataset, the first and second evaluation dataset and the at least one first calibration dataset, no second type of component is used, and wherein the second calibration dataset is used for training the artificial intelligence.
7. The method according to any of the preceding claims, characterized in that a third evaluation dataset of structure-borne noise data is each recorded using a second type of component, wherein the third evaluation dataset is evaluated after training the artificial intelligence using the second calibration dataset.
8. A method according to any one of the preceding claims, wherein the component is a part of a motor vehicle.
9. Method according to any of the preceding claims, characterized in that in evaluating the first, second and/or third evaluation data set, the movement of the respective component relative to the other component, the strength of the respective component and/or the presence of an imbalance is evaluated.
10. A system comprising a digital electronic storage medium and a digital electronic processing unit, wherein instructions are stored in the storage medium, wherein the processing unit is configured to facilitate reading out and executing the instructions, wherein the processing unit is configured to facilitate performing the method according to any of the preceding claims when executing the instructions.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE102022104304.5A DE102022104304A1 (en) | 2022-02-23 | 2022-02-23 | Computer-implemented method for evaluating structure-borne noise |
DE102022104304.5 | 2022-02-23 |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116644650A true CN116644650A (en) | 2023-08-25 |
Family
ID=87518725
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310055110.7A Pending CN116644650A (en) | 2022-02-23 | 2023-02-03 | Computer-implemented method for evaluating structure-borne noise |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN116644650A (en) |
DE (1) | DE102022104304A1 (en) |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102019003679A1 (en) | 2019-05-24 | 2019-11-07 | Daimler Ag | An engine analysis system for determining engine abnormality and methods for determining an engine abnormality |
DE102020206059A1 (en) | 2020-05-13 | 2021-11-18 | Siemens Healthcare Gmbh | Computer-implemented method and system for training an evaluation algorithm, computer program and electronically readable data carrier |
-
2022
- 2022-02-23 DE DE102022104304.5A patent/DE102022104304A1/en active Pending
-
2023
- 2023-02-03 CN CN202310055110.7A patent/CN116644650A/en active Pending
Also Published As
Publication number | Publication date |
---|---|
DE102022104304A1 (en) | 2023-08-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US6226760B1 (en) | Method and apparatus for detecting faults | |
KR101823746B1 (en) | Method for bearing fault diagnosis | |
CN107657250B (en) | Bearing fault detection and positioning method and detection and positioning model implementation system and method | |
CN108181105B (en) | Rolling bearing fault pre-diagnosis method and system based on logistic regression and J divergence | |
JP7102941B2 (en) | Information processing methods, information processing devices, and programs | |
CN115292834B (en) | Digital twin equipment fault diagnosis method, device and system | |
CN106441928A (en) | Method, device and system for vehicle fault detection | |
US20220101666A1 (en) | Method and Apparatus for Determining a Residual Service Life based on a Predictive Diagnosis of Components of an Electric Drive System Using Artificial Intelligence Techniques | |
CN111914883A (en) | Spindle bearing state evaluation method and device based on deep fusion network | |
CN112082639A (en) | Engine state diagnosis method and diagnosis modeling method thereof | |
CN114548152A (en) | Method for predicting residual life of marine sliding bearing based on transfer learning | |
CN116542984A (en) | Hardware defect detection method, device, computer equipment and storage medium | |
CN114462820A (en) | Bearing state monitoring and health management system performance testing and optimizing method and system | |
Meiners et al. | Towards an inline quality monitoring for crimping processes utilizing machine learning techniques | |
CN116644650A (en) | Computer-implemented method for evaluating structure-borne noise | |
CN114331214A (en) | Domain-adaptive bearing voiceprint fault diagnosis method and system based on reinforcement learning | |
CN117152119A (en) | Profile flaw visual detection method based on image processing | |
US20220385544A1 (en) | Early detection of cable failure in automotive networks | |
EP4095533A2 (en) | Early detection of cable failure in automotive networks | |
CN110956112A (en) | Novel high-reliability slewing bearing life evaluation method | |
CN116168720A (en) | Motor sound abnormality fault diagnosis method, system and storable medium | |
Decker et al. | Does your model think like an engineer? explainable ai for bearing fault detection with deep learning | |
CN113177484B (en) | Intelligent mechanical fault diagnosis method based on LOF self-encoding | |
US6947871B2 (en) | Yield/quality improvement using calculated failure rate derived from multiple component level parameters | |
CN115238904A (en) | Method, apparatus and computer program for creating training data in a vehicle |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |