WO2021046723A1 - 传感器模型的生成方法和系统、传感器测量方法和系统 - Google Patents

传感器模型的生成方法和系统、传感器测量方法和系统 Download PDF

Info

Publication number
WO2021046723A1
WO2021046723A1 PCT/CN2019/105197 CN2019105197W WO2021046723A1 WO 2021046723 A1 WO2021046723 A1 WO 2021046723A1 CN 2019105197 W CN2019105197 W CN 2019105197W WO 2021046723 A1 WO2021046723 A1 WO 2021046723A1
Authority
WO
WIPO (PCT)
Prior art keywords
sensor
model
target
sample
samples
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.)
Ceased
Application number
PCT/CN2019/105197
Other languages
English (en)
French (fr)
Inventor
白新
周晓舟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Siemens Ltd China
Siemens AG
Siemens Corp
Original Assignee
Siemens Ltd China
Siemens AG
Siemens Corp
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Siemens Ltd China, Siemens AG, Siemens Corp filed Critical Siemens Ltd China
Priority to EP19945109.7A priority Critical patent/EP4006767A4/en
Priority to PCT/CN2019/105197 priority patent/WO2021046723A1/zh
Priority to CN201980096752.3A priority patent/CN113874866A/zh
Publication of WO2021046723A1 publication Critical patent/WO2021046723A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0499—Feedforward networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/09—Supervised learning

Definitions

  • the present invention relates to the industrial field, in particular to a method and system for generating a sensor model, a method and system for sensor measurement, and a computer-readable storage medium.
  • FEA finite element analysis
  • CFD computational fluid dynamics
  • Baseline Simulation Model Baseline Simulation Model
  • a designer or engineer can iterate a series of prototypes/design concepts based on professional knowledge and using techniques such as finite element analysis (FEA) , Build a baseline simulation model, input material parameters, boundary conditions, geometric dimensions, and force points and other parameters, you can get the deformation value of each position of the C-shaped arm.
  • FEA finite element analysis
  • the predicted value based on the baseline simulation model is usually a relatively accurate value. Therefore, when the predicted deformation value meets the expected requirements, it can be determined that the C-arm design of the C-arm X-ray machine meets the design requirements and can be used at the factory. .
  • a method and system for generating a sensor model are proposed, and on the other hand, a method and system for sensor measurement and a computer-readable storage medium are proposed, which are used to replace physical sensors with sensor models. Sensors to monitor the true performance of the target device.
  • a method for generating a sensor model proposed in an embodiment of the present invention includes: determining at least one set of input parameter samples of a baseline simulation model; for each set of input parameter samples, inputting the input parameter samples into the baseline simulation model , Obtain a corresponding set of output parameters, select output parameters corresponding to a plurality of predetermined target measurement positions from the set of output parameters, as the output parameter samples of the plurality of target sensors; combine the input parameter samples with the The output parameter samples are stored as a set of samples in a sample set; for a single target sensor or a target sensor combination, the input parameter sample of each sample in the first number of samples in the sample set and the corresponding single target sensor or The output parameters of the target sensor combination train a constructed neural network model to obtain a temporary sensor model corresponding to the single target sensor or the target sensor combination; each of the second number of samples remaining in the sample set is used The input parameter samples of each sample and the output parameters corresponding to the single target sensor or the target sensor combination are used to verify the output accuracy of the sensor temporary model; for at
  • the method further includes: when the verification accuracy does not meet the setting requirements, returning to execute the step of determining at least one set of input parameter samples.
  • it further includes: the sample set obtained by training the initial sensor model of each single target sensor and/or target sensor combination is used as the target sample set; the single physical sensor or the physical sensor combination determined according to the set method needs to be arranged For at least one current target measurement location to be selected, for a single target sensor or target sensor combination of other target measurement locations, select the at least one current target measurement location corresponding to the output parameter sample of each sample in the target sample set At least one output parameter of the sample is added as the current auxiliary input parameter to the input parameter sample of the sample to form the current input parameter sample, and the current input parameter sample of each sample in the first number of samples in the target sample set and the corresponding input parameter sample
  • the output parameters of a single target sensor or the target sensor combination are trained on a constructed neural network model to obtain a sensor optimization model corresponding to the single target sensor or the target sensor combination; using the second quantity of the target sample set
  • the current input parameter sample of each sample in the sample and the output parameter corresponding to the single target sensor or the target sensor combination are used to verify the output
  • a sensor optimization model set consisting of at least one sensor optimization model of other target measurement locations is obtained; according to the sensor optimization model set obtained under each target measurement location to be selected, each sensor optimization model set can be obtained according to each target measurement location.
  • the verification accuracy of the sensor optimization model set selects the final at least one candidate target measurement position and the final sensor optimization model set.
  • a sensor model generation system proposed in an embodiment of the present invention includes: a first determination module for determining at least one set of input parameter samples of a baseline simulation model; a sample acquisition module for each set of input parameter samples, The input parameter samples are input into the baseline simulation model to obtain a corresponding set of output parameters, and the output parameters corresponding to a plurality of predetermined target measurement positions are selected from the set of output parameters as the output parameters of the plurality of target sensors.
  • Output parameter samples; the input parameter samples and the output parameter samples are stored as a set of samples in a sample collection; the first model training module is used for a single target sensor or a target sensor combination, using the sample collection.
  • the input parameter samples of each sample in the first number of samples and the output parameters corresponding to the single target sensor or the target sensor combination are trained on a constructed neural network model to obtain the corresponding single target sensor or the target sensor.
  • the sensor temporary model of the sensor combination using the input parameter sample of each sample in the second number of samples remaining in the sample set and the output parameter corresponding to the single target sensor or the target sensor combination to the sensor temporary model The output accuracy is verified; the model confirmation module is used for at least one temporary sensor model, and when the verification accuracy meets the first set requirement, each sensor temporary model is used as a corresponding single target sensor or a sensor combination of target sensors The initial model.
  • model confirmation module is further configured to notify the second determination module to execute the determination within the value range of the input parameters of each category when the verification accuracy does not meet the first setting requirement The operation of at least one set of input parameter samples.
  • it further includes: a second determination module, configured to determine the sample set of the initial sensor model obtained by training each single target sensor and/or target sensor combination as the target sample set; and the second model training module is configured to For at least one current target measurement location to be selected for a single physical sensor or combination of physical sensors determined in accordance with the setting method, a single target sensor or target sensor combination for other target measurement locations is determined from the target sample set.
  • a second determination module configured to determine the sample set of the initial sensor model obtained by training each single target sensor and/or target sensor combination as the target sample set
  • the second model training module is configured to For at least one current target measurement location to be selected for a single physical sensor or combination of physical sensors determined in accordance with the setting method, a single target sensor or target sensor combination for other target measurement locations is determined from the target sample set.
  • At least one output parameter corresponding to the at least one current target measurement position to be selected from the output parameter sample is selected as the current auxiliary input parameter and added to the input parameter sample of the sample to form the current input parameter sample, using the first in the target sample set
  • the current input parameter samples of each sample in the number of samples and the output parameters corresponding to the single target sensor or the target sensor combination are trained on a constructed neural network model to obtain the corresponding single target sensor or the target sensor A combined sensor optimization model; the output of the sensor optimization model using the current input parameter sample of each sample in the second number of samples in the target sample set and the output parameters corresponding to the single target sensor or the target sensor combination The accuracy is checked.
  • a sensor optimization model set consisting of at least one sensor optimization model of other target measurement positions is obtained; the decision module is used to optimize each sensor model set obtained under each target measurement position to be selected According to the verification accuracy of each sensor optimization model set, the final at least one candidate target measurement position and the final sensor optimization model set can be selected.
  • a sensor model generation system proposed in an embodiment of the present invention includes: at least one memory and at least one processor, wherein: the at least one memory is used to store a computer program; the at least one processor is used to call the at least one A computer program stored in a memory executes the method for generating a sensor model in any of the above-mentioned embodiments.
  • a computer-readable storage medium provided in an embodiment of the present invention has a computer program stored thereon; the computer program can be executed by a processor and implement the method for generating a sensor model in any of the above-mentioned embodiments.
  • the characteristics of the more accurate prediction value of the baseline simulation model are used to obtain a sample set that includes a large number of samples composed of input parameters and corresponding output parameters to train the required physical sensor.
  • the neural network model is used to obtain the sensor model (ie virtual sensor) corresponding to the required physical sensor, and then the sensor model can be used to replace individual physical sensors to realize the simulation measurement of the location where the target device is not suitable for installing the physical sensor; in addition, by using Using certain sensor models to replace some physical sensors can also save costs.
  • the output accuracy of the sensor model can be further improved. Therefore, in use, physical sensors can be set at some of the measurement positions, and the output, that is, the measured value, can be used as an auxiliary input parameter of the sensor model of other measurement positions to obtain a higher-precision output of the sensor model of other measurement positions.
  • Fig. 1 is an exemplary flowchart of a method for generating a sensor model in an embodiment of the present invention.
  • Fig. 2A is an exemplary flowchart of another method for generating a sensor model in an embodiment of the present invention.
  • Fig. 2B is a schematic diagram of the principle of the method shown in Fig. 2A.
  • 3A and 3B are schematic diagrams of the accuracy of the sensor optimization model set under different target measurement positions obtained by taking the design target as the displacement of the C-arm of the X-ray machine as an example in an example of the present invention.
  • Fig. 4 is a schematic diagram of the output comparison of the baseline simulation model, the sensor initial model, and the sensor optimization model under different target measurement positions in an example of the present invention.
  • Fig. 5 is an exemplary structure diagram of a generation system of a sensor model in an embodiment of the present invention
  • Fig. 6 is an exemplary structure diagram of another sensor model generation system in an embodiment of the present invention.
  • Fig. 7 is an exemplary structure diagram of a sensor measurement system in an embodiment of the present invention.
  • Fig. 8 is a schematic structural diagram of another sensor model generation system in an embodiment of the present invention.
  • the first determination module 502 Sample acquisition module 503
  • the first model training module 504 Model confirmation module 601
  • Second determination module 602 The second model training module 603 Decision module 710 Sensor model 711 Initial sensor model 712 Sensor optimization model 720 Physical sensor
  • the baseline simulation model replaces individual physical sensors.
  • the output of the baseline simulation model is a multi-predictor output, it is impossible to achieve individual output of the individual predictive values of interest, which is inconvenient to use; on the other hand, it is also because of the input and output of the baseline simulation model. It is impossible to achieve real-time data prediction.
  • the characteristic of the baseline simulation model's more accurate prediction value is considered to obtain a sample set including a large number of sample pairs composed of input parameters and corresponding output parameters to train a neural network model to obtain the corresponding data.
  • a sensor model of a physical sensor ie, a virtual sensor
  • the sensor model is used to replace individual physical sensors.
  • Fig. 1 is an exemplary flowchart of a method for generating a sensor model in an embodiment of the present invention. As shown in Figure 1, the method may include the following steps:
  • Step S11 Determine at least one set of input parameter samples of a baseline simulation model.
  • the baseline simulation model may have input parameter types determined according to the design goals of the target object, and value ranges of input parameters of each type determined according to various possible working conditions or empirical values.
  • the types of input parameters may include: material parameters, boundary conditions such as force points and force values, and parameters required for simulation.
  • at least one set of input parameter samples can be determined within the value range of each type of input parameter.
  • a set of values of the input parameters of each category can be randomly determined within their respective ranges to obtain a set of input parameter samples; and then a set of input parameters of each category can be randomly determined Take a value to get another set of input parameter samples.
  • Step S12 For each set of input parameter samples, input the input parameter samples into the baseline simulation model to obtain a corresponding set of output parameters, and select a plurality of predetermined target measurement positions from the set of output parameters.
  • the output parameters of are used as output parameter samples of a plurality of target sensors; the input parameter samples and the output parameter samples are stored as a set of samples in a sample collection.
  • the target measurement position refers to each potential position where the target sensor may need to be installed, and a plurality of target measurement positions can be determined according to the design target of the target object.
  • the design target of the target object is the deformation of the arm of the robot
  • the baseline simulation model will output a large number of output parameters.
  • These output parameters include not only the output parameters corresponding to the determined multiple target measurement positions, but also many other uninteresting Therefore, in this step, only the output parameters corresponding to the multiple target measurement positions can be selected as the output parameter samples corresponding to the input parameter samples, and the selected output parameter samples are the corresponding target sensors Sample of output parameters.
  • Step S13 For a single target sensor or a target sensor combination, the input parameter sample of each sample in the first number of samples in the sample set and the output parameter corresponding to the single target sensor or the target sensor combination are used.
  • the constructed neural network model is trained to obtain a temporary sensor model corresponding to the single target sensor or the target sensor combination; the input parameter sample of each sample in the second number of samples remaining in the sample set and the corresponding
  • the output parameters of a single target sensor or a combination of the target sensors verify the output accuracy of the temporary model of the sensor. Normally, the first quantity is much larger than the second quantity.
  • the output parameter used for training and accuracy verification needs to be the output parameter corresponding to the target sensor in the output parameter sample, that is, the target position where the target sensor needs to be installed
  • the corresponding output parameter if the sensor model is trained for a target sensor combination, such as two target sensors corresponding to different target positions, the output parameter used for training and accuracy verification needs to be the target sensor combination in the output parameter sample.
  • Step S14 For at least one temporary sensor model, determine whether the verification accuracy meets the set requirements, and if so, perform step S15; otherwise, return to perform step S11 to further expand the number of samples in the sample set.
  • the calibration accuracy of the temporary sensor model obtained after training based on the same sample set may be different, for example, some have higher accuracy, and some have higher accuracy. Slightly lower, so the specific criteria for confirming whether the training is completed can have many situations. For example, it can be confirmed that the training is completed when the temporary model of each sensor meets the accuracy requirements; it can also be confirmed that the training is completed when most of the temporary sensor models meet the accuracy requirements; or, it can also be determined according to other calculation rules to determine the temporary training of each sensor. The accuracy requirements of the model.
  • step S15 the temporary model of each sensor is used as the corresponding initial sensor model of a single target sensor or a combination of target sensors.
  • either a single target sensor or an initial sensor model of a target sensor combination can be used in actual applications to replace physical sensors that cannot be installed at the corresponding location or reduce the use of installed physical sensors in order to save costs.
  • the initial models of these sensors can also be used for parameter prediction and product evaluation during the product design stage.
  • the measurement values of physical sensors at different positions of the same object may be related to some extent.
  • the measurement values of physical sensors at adjacent measurement positions should be It will be relatively close. Therefore, in actual applications, physical sensors can be placed in some measurement locations, and sensor models can be used in other locations to simulate, and the measured values of the physical sensors can be used as auxiliary input parameters of the relevant sensor model.
  • the sensor optimization model with auxiliary input parameters can be further trained, and the auxiliary input parameters in the training phase can be the parameters corresponding to the measurement position output by the baseline simulation model.
  • the sample set of the initial model of each sensor can be obtained by training for training.
  • an exemplary flowchart of another method for generating a sensor model in another embodiment may be specifically shown in FIG. 2A, including the following steps:
  • the output parameters of the sensor are used to verify the output accuracy of the sensor optimization model.
  • each current target measurement position to be selected or the current target measurement position combination at least one sensor optimization model of other target measurement positions can be obtained.
  • each current target measurement position to be selected Or at least one sensor optimization model corresponding to the current target measurement position combination to be selected is called a sensor optimization model set.
  • the final at least one target measurement position to be selected can be selected according to the verification accuracy of each sensor optimization model set (for example, a single target measurement position to be selected or a target measurement position to be selected). Select the target measurement position combination) and the final sensor optimization model set.
  • FIG. 3A and FIG. 3B respectively show the accuracy schematic diagrams of the sensor optimization model set under different target measurement positions obtained by taking the design target as the displacement of the C-arm of the X-ray machine as an example.
  • the best position of the physical sensor on the C-arm X-ray machine with three degrees of freedom is determined by using the above-mentioned method in the embodiment of the present application.
  • the output parameters of the baseline simulation model corresponding to different target measurement positions in the same target sample set are used as auxiliary input parameters to train the sensor model.
  • the two figures respectively show the predictions from the sensor optimization model and the baseline simulation model corresponding to different target measurement positions.
  • the triangular point at the center is the output of the baseline simulation model (FEA), which is also That is, the accuracy calibration reference point, two curves (U line and D line ) constitute the accuracy error range, and the dot is the output of the sensor optimization model (SOM) set.
  • FEA baseline simulation model
  • SOM sensor optimization model
  • the principle of the method shown in FIG. 2A can be as shown in FIG. 2B.
  • the candidate target measurement positions 321, 322, ... 32n are sequentially determined, based on the candidate target measurement positions 321, 321, and 321 determined each time.
  • a sensor optimization model set 331, 332,...33n can be obtained, and then according to the comprehensive evaluation, the best target measurement position and its corresponding sensor optimization model set 340 can be obtained.
  • the target measurement position to be selected determined each time may be a single target measurement position or a combination of target measurement positions.
  • physical sensors can be arranged at the corresponding target measurement positions to be selected according to the actual situation, and sensor optimization models can be used for simulation measurements at other target measurement positions, and the measured values of the physical sensors, namely output parameters, can be used As an auxiliary input parameter of each sensor optimization model.
  • Figure 4 shows the output comparison schematic diagram of the baseline simulation model, the sensor initial model and the sensor optimization model under different target measurement positions.
  • the dots are the output of the baseline simulation model (FEA), and the pentagonal points are the sensors.
  • the output of the initial model (SIM) is the output of the sensor optimization model (SOM).
  • the output of the sensor optimization model is relatively close to the output of the baseline simulation model from beginning to end, while the output of the sensor initial model has a beginning Gradually deviate from the output of the baseline simulation model.
  • the output accuracy of the sensor optimization model based on the output of the physical sensor as an auxiliary input parameter is better than the output accuracy of the initial sensor model.
  • the physical sensor, the sensor optimization model, and the sensor initial model can also be used in combination; or only the sensor initial model and the physical sensor can be used at the same time, but the measured value of the physical sensor is not used as an auxiliary input parameter of the sensor initial model ; Or, it is also possible to use only the initial model of the sensor.
  • an embodiment of the present invention also proposes a sensor measurement method, which may include the following steps: use N sensor models to perform analog detection at M1 measurement positions among the M measurement positions. The remaining M2 measurement positions among the M measurement positions are actually detected using M2 physical sensors.
  • M M1+M2, M1 is greater than or equal to N; and M, M1, and N are integers greater than or equal to 1, and M2 is an integer greater than or equal to 0.
  • the sensor model generation method and sensor measurement method in the embodiment of the present invention have been described in detail above, and the sensor model generation system and the sensor measurement system in the embodiment of the present invention will be described in detail below.
  • the sensor model generation system in the embodiment of the present invention can be used to implement the sensor model generation method in the embodiment of the present invention.
  • the sensor measurement system in the embodiment of the present invention can be used to implement the sensor measurement method in the embodiment of the present invention.
  • Fig. 5 is an exemplary structure diagram of a generation system of a sensor model in an embodiment of the present invention. As shown in FIG. 5, the system may include: a first determination module 501, a sample acquisition module 502, a first model training module 503, and a model confirmation module 504.
  • the first determining module 501 is used to determine at least one set of input parameter samples of a baseline simulation model.
  • the sample acquisition module 502 is configured to input the input parameter samples into the baseline simulation model for each set of input parameter samples to obtain a corresponding set of output parameters, and select a plurality of corresponding predetermined output parameters from the set of output parameters.
  • the output parameters of the target measurement position are used as output parameter samples of a plurality of target sensors; the input parameter samples and the output parameter samples are stored as a set of samples in a sample set.
  • the first model training module 503 is used for a single target sensor or a target sensor combination, using the input parameter sample of each sample in the first number of samples in the sample set and the corresponding single target sensor or the target sensor combination Train a constructed neural network model to obtain a temporary sensor model corresponding to the single target sensor or the target sensor combination; use the input parameters of each sample in the second number of samples remaining in the sample set The sample and the output parameters corresponding to the single target sensor or the target sensor combination are used to verify the output accuracy of the temporary model of the sensor.
  • the model confirmation module 504 is configured to, for at least one temporary sensor model, when the verification accuracy meets the first set requirement, use each temporary sensor model as the corresponding single target sensor or the initial sensor model of the target sensor combination. Further, the model confirmation module 504 is further configured to notify the first determination module 501 to perform the operation of determining at least one set of input parameter samples of the baseline simulation model when the verification accuracy does not meet the first setting requirement.
  • Fig. 6 is an exemplary structure diagram of another sensor model generation system in an embodiment of the present invention. As shown in FIG. 6, the system may include: a second determination module 601, a second model training module 602, and a decision module 603.
  • the second determining module 601 is configured to determine the sample set of the sensor initial model of each single target sensor and/or target sensor combination trained in the embodiment shown in FIG. 5 as the target sample set.
  • the second model training module 602 is used to determine at least one current target measurement location to be selected for a single physical sensor or a combination of physical sensors determined in a set manner, and for a single target sensor or target sensor combination of other target measurement locations, from all Selecting at least one output parameter corresponding to the at least one current target measurement position to be selected from the output parameter sample of each sample in the target sample set as the current auxiliary input parameter and adding it to the input parameter sample of the sample to form a current input parameter sample, Use the current input parameter sample of each sample in the first number of samples in the target sample set and the output parameter corresponding to the single target sensor or the target sensor combination to train a constructed neural network model to obtain the corresponding A sensor optimization model of a single target sensor or the target sensor combination; using the current input parameter sample of each sample in the second number of samples in the target sample set and the output parameter corresponding to the single target sensor or the target sensor combination The output accuracy of the sensor optimization model is verified. For at least one current target measurement position to be selected, a sensor optimization model set composed of at
  • the decision module 603 is used to select the final at least one target measurement location to be selected and the final sensor optimization model set according to the verification accuracy of each sensor optimization model set according to the sensor optimization model set obtained under each target measurement location to be selected. .
  • the sensor model generation system shown in FIG. 5 and FIG. 6 can be implemented in different devices, or can also be implemented in the same device.
  • Fig. 7 is an exemplary structure diagram of a sensor measurement system in an embodiment of the present invention. As shown in FIG. 7, the system may include N sensor models 710 and M2 physical sensors 720.
  • the N sensor models 710 are used to perform analog detection on M1 measurement positions among the M measurement positions.
  • the M2 physical sensors 720 are used to actually detect the remaining M2 measurement positions among the M measurement positions.
  • M M1+M2, M1 is greater than or equal to N; and M, M1, and N are integers greater than or equal to 1, and M2 is an integer greater than or equal to 0.
  • FIG. 8 is a schematic structural diagram of another sensor model generation system in an embodiment of the present invention. As shown in FIG. 8, the system may include: at least one memory 81 and at least one processor 82. In addition, some other components may also be included, such as communication ports. These components communicate through the bus 83.
  • At least one memory 81 is used to store a computer program.
  • the computer program can be understood as including various modules of the sensor model generation system shown in FIG. 5 and/or FIG. 6.
  • at least one memory 81 may also store an operating system and the like.
  • Operating systems include but are not limited to: Android operating system, Symbian operating system, Windows operating system, Linux operating system and so on.
  • the at least one processor 82 is configured to call at least one computer program stored in the memory 81 to execute the method for generating a sensor model described in the embodiment of the present invention.
  • the processor 82 may be a CPU, a processing unit/module, an ASIC, a logic module or a programmable gate array, etc. It can receive and send data through the communication port.
  • a hardware module may include specially designed permanent circuits or logic devices (such as dedicated processors, such as FPGAs or ASICs) to complete specific operations.
  • the hardware module may also include programmable logic devices or circuits temporarily configured by software (for example, including general-purpose processors or other programmable processors) for performing specific operations.
  • software for example, including general-purpose processors or other programmable processors
  • the embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored, and the computer program can be executed by a processor and implement the method for generating the sensor model described in the embodiment of the present invention.
  • a system or device equipped with a storage medium may be provided, and the software program code for realizing the function of any one of the above-mentioned embodiments is stored on the storage medium, and the computer (or CPU or MPU of the system or device) ) Read and execute the program code stored in the storage medium.
  • an operating system or the like operating on the computer can also be used to complete part or all of the actual operations through instructions based on the program code.
  • Implementations of storage media used to provide program codes include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), Magnetic tape, non-volatile memory card and ROM.
  • the program code can be downloaded from the server computer via a communication network.
  • the characteristics of the more accurate prediction value of the baseline simulation model are used to obtain a sample set that includes a large number of samples composed of input parameters and corresponding output parameters to train the required physical sensor.
  • the neural network model is used to obtain the sensor model (ie virtual sensor) corresponding to the required physical sensor, and then the sensor model can be used to replace individual physical sensors to realize the simulation measurement of the location where the target device is not suitable for installing the physical sensor; in addition, by using Using certain sensor models to replace some physical sensors can also save costs.
  • the output accuracy of the sensor model can be further improved. Therefore, in use, physical sensors can be set at some of the measurement positions, and the output, that is, the measured value, can be used as an auxiliary input parameter of the sensor model of other measurement positions to obtain a higher-precision output of the sensor model of other measurement positions.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Testing Or Calibration Of Command Recording Devices (AREA)

Abstract

一种传感器模型的生成方法和系统、传感器测量方法和系统、以及存储介质。其中，生成方法包括：预先确定一基线仿真模型的至少一组输入参数样本（S11），并得到针对每组输入参数样本的一组输出参数样本，将其作为一组样本存储到一样本集中（S12）；针对一目标传感器或目标传感器组合，利用样本集中第一数量的样本对一构建的神经网络模型进行训练，得到对应的传感器临时模型；利用样本集中剩余的样本对传感器临时模型的输出精度进行校验（S13）；在校验精度满足设定要求时，将每个传感器临时模型作为所对应的单个目标传感器或目标传感器组合的传感器初始模型（S14-S15），该传感器模型的生成方法和系统利用传感器模型来代替物理传感器来监测目标设备的真实性能。

Description

传感器模型的生成方法和系统、传感器测量方法和系统 技术领域
本发明涉及工业领域，特别是一种传感器模型的生成方法和系统、传感器测量方法和系统、以及计算机可读存储介质。
背景技术
在工业领域中，为了在设计阶段对特定领域的问题做出准确的预测，如评估目标设备的设计性能等，通常需要利用有限元分析(FEA)、计算流体动力学(CFD)等技术建立一基线仿真模型(Baseline Simulation Model)，并利用该基线仿真模型进行模拟仿真，输入各工况条件下的相关参数，得到对应的性能等参数的输出。在输出满足预期要求时，则确定该目标设备的设计符合要求。
例如，以预测C形臂X光机的C形臂各位置的形变为例，在设计阶段，可由设计师或工程师基于专业知识，利用有限元分析(FEA)等技术迭代一系列原型/设计概念，搭建出基线仿真模型，通过输入材料参数、边界条件、几何尺寸、以及受力点等参数，便可得到C形臂各个位置的形变值。基于基线仿真模型得到的预测值通常是比较准确的值，因此在预测出的形变值满足预期要求时，便可确定该C形臂X光机的C形臂设计满足设计要求，从而可以出厂使用。
在出厂后的使用阶段，要想对目标设备的真实性能进行监测和评估，则需要依赖于安装在现场各位置的物理传感器的实际测量值或通过经验模型的推导值。
发明内容
有鉴于此，本发明实施例中一方面提出了一种传感器模型的生成方法和系统，另一方面提出了一种传感器测量方法和系统以及计算机可读存储介质，用于利用传感器模型来代替物理传感器来监测目标设备的真实性能。
本发明实施例中提出的一种传感器模型的生成方法，包括：确定一基线仿真模型的至少一组输入参数样本；针对每组输入参数样本，将所述输入参数样本输入所述基线仿真模型中，得到对应的一组输出参数，从所述一组输出参数中选取对应预先确定的复数 个目标测量位置的输出参数，作为复数个目标传感器的输出参数样本；将所述输入参数样本和所述输出参数样本作为一组样本存储到一样本集中；针对一单个目标传感器或一目标传感器组合，利用所述样本集中第一数量的样本中每个样本的输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对一构建的神经网络模型进行训练，得到对应所述单个目标传感器或所述目标传感器组合的传感器临时模型；利用所述样本集中剩余的第二数量的样本中每个样本的输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对所述传感器临时模型的输出精度进行校验；针对至少一个传感器临时模型，在校验精度满足设定要求时，将每个传感器临时模型作为所对应的单个目标传感器或目标传感器组合的传感器初始模型。
在一个实施方式中，进一步包括：在校验精度不满足所述设定要求时，返回执行所述确定至少一组输入参数样本的步骤。
在一个实施方式中，进一步包括：将训练得到各单个目标传感器和/或目标传感器组合的传感器初始模型的样本集作为目标样本集；针对按照设定方式确定的单个物理传感器或物理传感器组合需布置的至少一个当前待选目标测量位置，对其他目标测量位置的单个目标传感器或目标传感器组合，从所述目标样本集中每个样本的输出参数样本中选取所述至少一个当前待选目标测量位置对应的至少一个输出参数作为当前辅助输入参数加入所述样本的输入参数样本中，构成当前输入参数样本，利用所述目标样本集中第一数量的样本中每个样本的当前输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对一构建的神经网络模型进行训练，得到对应所述单个目标传感器或所述目标传感器组合的传感器优化模型；利用所述目标样本集中第二数量的样本中每个样本的当前输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对所述传感器优化模型的输出精度进行校验。针对至少一个当前待选目标测量位置，得到由其他目标测量位置的至少一个传感器优化模型构成的一个传感器优化模型集；根据各待选目标测量位置下得到的各传感器优化模型集，可根据每个传感器优化模型集的校验精度选取最终的至少一个待选目标测量位置和最终的传感器优化模型集。
本发明实施例中提出的一种传感器测量方法，包括：在M个测量位置中的M1个测量位置利用N个传感器模型进行模拟检测；在M个测量位置中剩余的M2个测量位置利用M2个物理传感器进行实际检测；其中，M＝M1+M2，M1大于或等于N；且M、M1和N分别为大于或等于1的整数，M2为大于或等于0的整数；所述N个传感器包括N1个传感器初始模型和N2个以所述M2个物理传感器的输出值为辅助输入参数的传感器优化模型；其中，N＝N1+N2，N1和N2分别为大于或等于0的整数，且二者不同时为0； 述传感器初始模型根据权利要求1或2中所述的传感器模型的生成方法得到；所述传感器优化模型和所述M2个物理传感器的测量位置根据权利要求3中所述的传感器模型的生成方法得到。
本发明实施例中提出的一种传感器模型的生成系统，包括：第一确定模块，用于确定一基线仿真模型的至少一组输入参数样本；样本获取模块，用于针对每组输入参数样本，将所述输入参数样本输入所述基线仿真模型中，得到对应的一组输出参数，从所述一组输出参数中选取对应预先确定的复数个目标测量位置的输出参数，作为复数个目标传感器的输出参数样本；将所述输入参数样本和所述输出参数样本作为一组样本存储到一样本集中；第一模型训练模块，用于针对一单个目标传感器或一目标传感器组合，利用所述样本集中第一数量的样本中每个样本的输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对一构建的神经网络模型进行训练，得到对应所述单个目标传感器或所述目标传感器组合的传感器临时模型；利用所述样本集中剩余的第二数量的样本中每个样本的输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对所述传感器临时模型的输出精度进行校验；模型确认模块，用于针对至少一个传感器临时模型，在校验精度满足第一设定要求时，将每个传感器临时模型作为所对应的单个目标传感器或目标传感器组合的传感器初始模型。
在一个实施方式中，所述模型确认模块进一步用于在校验精度不满足第一设定要求时，通知所述第二确定模块执行所述在各类别的输入参数的取值范围内，确定至少一组输入参数样本的操作。
在一个实施方式中，进一步包括：第二确定模块，用于将训练得到各单个目标传感器和/或目标传感器组合的传感器初始模型的样本集确定为目标样本集；第二模型训练模块，用于针对按照设定方式确定的单个物理传感器或物理传感器组合需布置的至少一个当前待选目标测量位置，对其他目标测量位置的单个目标传感器或目标传感器组合，从所述目标样本集中每个样本的输出参数样本中选取所述至少一个当前待选目标测量位置对应的至少一个输出参数作为当前辅助输入参数加入所述样本的输入参数样本中，构成当前输入参数样本，利用所述目标样本集中第一数量的样本中每个样本的当前输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对一构建的神经网络模型进行训练，得到对应所述单个目标传感器或所述目标传感器组合的传感器优化模型；利用所述目标样本集中第二数量的样本中每个样本的当前输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对所述传感器优化模型的输出精度进行校验。针对至少一个当前待选目标测量位置，得到由其他目标测量位置的至少一个传感器 优化模型构成的一个传感器优化模型集；决策模块，用于根据各待选目标测量位置下得到的各传感器优化模型集，可根据每个传感器优化模型集的校验精度选取最终的至少一个待选目标测量位置和最终的传感器优化模型集。
本发明实施例中提出的一种传感器测量系统，包括：N个传感器模型，用于对M个测量位置中的M1个测量位置进行模拟检测；和M2个物理传感器，用于对M个测量位置中剩余的M2个测量位置进行实际检测；其中，M＝M1+M2，M1大于或等于N；且M、M1和N分别为大于或等于1的整数，M2为大于或等于0的整数；所述N个传感器模型包括N1个传感器初始模型和N2个以所述M2个物理传感器的输出值为辅助输入参数的传感器优化模型；其中，N＝N1+N2，N1和N2分别为大于或等于0的整数，且二者不同时为0；所述传感器初始模型由权利要求5或6中所述的传感器模型的生成系统得到；所述传感器优化模型和所述M2个物理传感器的测量位置由权利要求3中所述的传感器模型的生成系统得到。
本发明实施例中提出的一种传感器模型的生成系统，包括：至少一个存储器和至少一个处理器，其中：所述至少一个存储器用于存储计算机程序；所述至少一个处理器用于调用所述至少一个存储器中存储的计算机程序，执行如上所述的任一实施方式中的传感器模型的生成方法。
本发明实施例中提出的一种计算机可读存储介质，其上存储有计算机程序；所述计算机程序能够被一处理器执行并实现如上所述的任一实施方式中的传感器模型的生成方法。
从上述方案中可以看出，由于本发明实施例中利用基线仿真模型预测值较准确的特点来获取包括大量的由输入参数和对应的输出参数构成的样本的样本集来训练所需物理传感器的神经网络模型，得到对应所需物理传感器的传感器模型(即虚拟传感器)，之后便可利用该传感器模型来代替个别物理传感器，以实现目标设备不适合安装物理传感器的位置的仿真测量；此外通过利用某些传感器模型来代替一些物理传感器，还可以节约成本。
此外，通过利用某些测量位置对应的基线仿真模型输出来作为其它测量位置的辅助输入参数来训练其它测量位置的传感器模型，可进一步提高传感器模型的输出精度。从而在使用中可在所述某些测量位置设置物理传感器，并利用其输出即测量值作为其它测量位置的传感器模型的辅助输入参数，得到其它测量位置的传感器模型的较高精度的输出。
另外，通过将物理传感器、传感器优化模型和传感器初始模型灵活搭配使用，可以 应对各种应用场景，提高了目标设备的整体测量精度。
附图说明
下面将通过参照附图详细描述本发明的优选实施例，使本领域的普通技术人员更清楚本发明的上述及其它特征和优点，附图中：
图1为本发明实施例中一种传感器模型的生成方法的示例性流程图。
图2A为本发明实施例中另一种传感器模型的生成方法的示例性流程图。
图2B为图2A所示方法的原理示意图。
图3A和图3B为本发明一个例子中以设计目标为X光机C形臂的位移为例得到的不同待选目标测量位置下的传感器优化模型集的精度示意图。
图4为本发明一个例子中基线仿真模型、传感器初始模型和传感器优化模型三者在不同目标测量位置下的输出对比示意图。
图5为本发明实施例中一种传感器模型的生成系统的示例性结构图
图6为本发明实施例中另一种传感器模型的生成系统的示例性结构图。
图7为本发明实施例中传感器测量系统的示例性结构图。
图8为本发明实施例中又一种传感器模型的生成系统的结构示意图。
其中，附图标记如下：
标号 含义
S11～S15、S21～S23 步骤
501 第一确定模块
502 样本获取模块
503 第一模型训练模块
504 模型确认模块
601 第二确定模块
602 第二模型训练模块
603 决策模块
710 传感器模型
711 传感器初始模型
712 传感器优化模型
720 物理传感器
81 存储器
82 处理器
83 总线
具体实施方式
为了描述上的简洁和直观，下文通过描述若干代表性的实施方式来对本发明的方案进行阐述。实施方式中大量的细节仅用于帮助理解本发明的方案。但是很明显，本发明的技术方案实现时可以不局限于这些细节。为了避免不必要地模糊了本发明的方案，一些实施方式没有进行细致地描述，而是仅给出了框架。下文中，“包括”是指“包括但不限于”，“根据……”是指“至少根据……，但不限于仅根据……”。由于汉语的语言习惯，下文中没有特别指出一个成分的数量时，意味着该成分可以是一个也可以是多个，或可理解为至少一个。
本发明实施例中，考虑到实际应用中，目标设备的有些位置可能不适合安装物理传感器，或者基于成本或空间等方面的考虑没办法在各个目标位置都安装物理传感器，但此时又无法用基线仿真模型来代替个别的物理传感器，一方面因为基线仿真模型的输出是多预测值输出，无法实现个别感兴趣预测值的单独输出，使用不方便；另一方面也是因为基线仿真模型的输入输出无法实现实时数据的预测。为此，本发明实施例中，考虑利用基线仿真模型预测值较准确的特点来获取包括大量的由输入参数和对应的输出参数构成的样本对的样本集来训练一神经网络模型，得到对应所需物理传感器的传感器模型(即虚拟传感器)，之后利用该传感器模型来代替个别物理传感器。
为了使本发明的技术方案及优点更加清楚明白，以下结合附图及实施方式，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施方式仅仅用以阐述性说明本发明，并不用于限定本发明的保护范围。
图1为本发明实施例中一种传感器模型的生成方法的示例性流程图。如图1所示，该方法可包括如下步骤：
步骤S11、确定一基线仿真模型的至少一组输入参数样本。
本步骤中，基线仿真模型可具有根据目标对象的设计目标确定的输入参数的类别，以及根据各种可能的工况条件或经验值等确定的各类别输入参数的取值范围。例如，目标对象的设计目标为机器人的手臂形变，则输入参数的类别可包括：材料参数、受力点和受力值等边界条件、以及仿真所需的参数等。相应地，至少一组输入参数样本可在各 类别的输入参数的取值范围内确定。例如，可根据各类别的输入参数的取值范围，在各自的范围内随机确定各类别的输入参数的一组取值，得到一组输入参数样本；再随机确定各类别的输入参数的一组取值，得到另一组输入参数样本。
步骤S12、针对每组输入参数样本，将所述输入参数样本输入所述基线仿真模型中，得到对应的一组输出参数，从所述一组输出参数中选取对应预先确定的复数个目标测量位置的输出参数，作为复数个目标传感器的输出参数样本；将所述输入参数样本和所述输出参数样本作为一组样本存储到一样本集中。
其中，目标测量位置指的是可能需要安装目标传感器的各潜在位置，复数个目标测量位置可根据目标对象的设计目标确定。例如，对于目标对象的设计目标为机器人的手臂形变的情况，可能需要布置目标传感器的复数个目标测量位置可包括沿机器人手臂分布的多个测量位置，例如沿机器人手臂分布的几个、十几个甚至数十个测量位置等。
实际应用中，由于对应每组输入参数样本，基线仿真模型会输出大量的输出参数，这些输出参数除了包括所确定的复数个目标测量位置所对应的输出参数以外，还包括很多其他的非感兴趣点的输出参数，因此，本步骤中，可仅从中选取复数个目标测量位置所对应的输出参数作为输入参数样本对应的输出参数样本即可，且所选取的输出参数样本即为对应的目标传感器的输出参数样本。
步骤S13、针对一单个目标传感器或一目标传感器组合，利用所述样本集中第一数量的样本中每个样本的输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对一构建的神经网络模型进行训练，得到对应所述单个目标传感器或所述目标传感器组合的传感器临时模型；利用所述样本集中剩余的第二数量的样本中每个样本的输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对所述传感器临时模型的输出精度进行校验。通常情况下，第一数量远大于第二数量。
本步骤中，若仅针对单个的目标传感器训练其传感器模型，则用于训练以及精度校验的输出参数需为输出参数样本中对应该目标传感器的输出参数，即需要安装该目标传感器的目标位置所对应的输出参数；若针对一个目标传感器组合如对应不同目标位置的两个目标传感器训练其传感器模型，则用于训练以及精度校验的输出参数需为输出参数样本中对应该目标传感器组合即对应不同目标位置的两个目标传感器的输出参数。
步骤S14、针对至少一个传感器临时模型，判断校验精度是否满足设定要求，如果是，则执行步骤S15；否则，返回执行步骤S11，以进一步扩大样本集中的样本数量。
本步骤中，考虑到要训练的传感器模型可能不只一个，相应地，基于相同的样本集训练后得到的传感器临时模型的校验精度可能也不尽相同，例如有的精度较高，有的精 度稍低一些，因此具体确认是否训练完成的标准可以有多种情况。例如，可以是每个传感器临时模型都满足精度要求时，确认训练完成；也可以是大多数传感器临时模型满足精度要求时，确认训练完成；或者，也可以是根据其他计算规则，确定各传感器临时模型的精度要求。
步骤S15，将每个传感器临时模型作为所对应的单个目标传感器或目标传感器组合的传感器初始模型。
其中，无论是单个目标传感器还是目标传感器组合的传感器初始模型均可以在实际应用中替代无法在对应位置安装的物理传感器使用或为了节约成本而减少安装的物理传感器使用。此外，这些传感器初始模型也可以在产品的设计阶段进行参数的预测以及产品的评估等。
进一步地，考虑到现实应用中，同一对象不同位置处的物理传感器的测量值可能会有某种程度的关联，例如针对机器人手臂的受力形变，相邻测量位置处的物理传感器的测量值应该会比较接近，因此实际应用中，可在部分测量位置布置物理传感器，在其他位置采用传感器模型来仿真，并利用物理传感器的测量值作为相关传感器模型的辅助输入参数。基于此，本发明实施例中，可进一步训练带有辅助输入参数的传感器优化模型，并且训练阶段的辅助输入参数可采用基线仿真模型输出的对应测量位置的参数。
具体实现时，可利用训练得到各传感器初始模型的样本集进行训练。例如，具体可如图2A所示的又一实施方式中的另一种传感器模型的生成方法的示例性流程图，包括如下步骤：
S21、将图1所示实施例中训练得到各单个目标传感器和/或目标传感器组合的传感器初始模型的样本集作为目标样本集。
S22、针对按照设定方式如轮询方式或枚举方式等确定的单个物理传感器或物理传感器组合需布置的至少一个当前待选目标测量位置，对其他目标测量位置的单个目标传感器或目标传感器组合，从所述目标样本集中每个样本的输出参数样本中选取所述至少一个当前待选目标测量位置对应的至少一个输出参数作为当前辅助输入参数加入所述样本的输入参数样本中，构成当前输入参数样本，利用所述目标样本集中第一数量的样本中每个样本的当前输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对一构建的神经网络模型进行训练，得到对应所述单个目标传感器或所述目标传感器组合的传感器优化模型；利用所述目标样本集中第二数量的样本中每个样本的当前输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对所述传感器优化模型的输出精度进行校验。针对至少一个当前待选目标测量位置，得到由其他目标 测量位置的至少一个传感器优化模型构成的一个传感器优化模型集。
执行本步骤S22后，针对每个当前待选目标测量位置或当前待选目标测量位置组合，可得到其他目标测量位置的至少一个传感器优化模型，为简便起见，将每个当前待选目标测量位置或当前待选目标测量位置组合对应的至少一个传感器优化模型称为一个传感器优化模型集。
S23、根据各待选目标测量位置下得到的各传感器优化模型集，可根据每个传感器优化模型集的校验精度选取最终的至少一个待选目标测量位置(如单个待选目标测量位置或待选目标测量位置组合)和最终的传感器优化模型集。
例如，图3A和图3B分别示出了以设计目标为X光机C形臂的位移为例得到的不同待选目标测量位置下的传感器优化模型集的精度示意图。利用本申请实施例中的上述方法确定了具有三个自由度的C形臂X光机上物理传感器的最佳位置。在每个图中，使用同一目标样本集中不同待选目标测量位置对应的基线仿真模型的输出参数作为辅助输入参数来训练传感器模型。如图3A和图3B所示，两个图中分别示出了对应不同目标测量位置的来自传感器优化模型和基线仿真模型的预测，位于中心的三角点为基线仿真模型(FEA)的输出，也即精度校验基准点，两条曲线(U line和D line)构成了精度的误差范围，而圆点为传感器优化模型(SOM)集的输出。通过将图3A和图3B进行对比，可以发现图3A中的传感器优化模型集的输出几乎均在精度误差范围内，而图3B中的传感器模型集的输出则有好几个超出了精度误差范围，因此图3A中的传感器优化模型集在评估位移分布方面表现出更好的性能，这同时也表明，当与相应的传感器优化模型一起使用时，在图3A对应的待选目标测量位置处放置物理传感器(如果需要)可用于提供更准确的预测。
可见，图2A所示方法的原理可如图2B所示，基于目标样本集310，依次确定各待选目标测量位置321、322、……32n，基于每次确定的待选目标测量位置321、322、……32n，可得到一个传感器优化模型集331、332、……33n，之后根据综合评估，得到最佳的目标测量位置及其对应的传感器优化模型集340。其中，每次确定的待选目标测量位置可以是一个单个的目标测量位置，也可以是一个目标测量位置组合。
在随后的应用中，可根据实际情况在对应的待选目标测量位置处布置物理传感器，在其他的目标测量位置采用传感器优化模型来进行仿真测量，并利用所述物理传感器的测量值即输出参数作为各传感器优化模型的辅助输入参数。
图4中示出了基线仿真模型、传感器初始模型和传感器优化模型三者在不同目标测量位置下的输出对比示意图，其中，圆点为基线仿真模型(FEA)的输出，五边形点为 传感器初始模型(SIM)的输出，三角点为传感器优化模型(SOM)的输出，可以看出，传感器优化模型的输出自始至终都比较接近于基线仿真模型的输出，而传感器初始模型的输出则有一段开始逐渐偏离基线仿真模型的输出。也就是说，基于物理传感器的输出作为辅助输入参数的传感器优化模型的输出精度要优于传感器初始模型的输出精度。
当然，其他实施方式中，也可以物理传感器、传感器优化模型和传感器初始模型混合使用；或者也可以仅同时使用传感器初始模型和物理传感器，只是物理传感器的测量值不作为传感器初始模型的辅助输入参数；又或者，也可仅使用传感器初始模型。
例如，本发明实施例中还提出一种传感器测量方法，其可包括如下步骤：在M个测量位置中的M1个测量位置利用N个传感器模型进行模拟检测。在M个测量位置中剩余的M2个测量位置利用M2个物理传感器进行实际检测。其中，M＝M1+M2，M1大于或等于N；且M、M1和N分别为大于或等于1的整数，M2为大于或等于0的整数。所述N个传感器包括N1个传感器初始模型和N2个以所述M2个物理传感器的输出值为辅助输入参数的传感器优化模型；其中，N＝N1+N2，N1和N2分别为大于或等于0的整数，且二者不同时为0。
以上对本发明实施例中的传感器模型的生成方法及传感器测量方法进行了详细描述，下面再对本发明实施例中的传感器模型的生成系统及传感器测量系统进行详细介绍。本发明实施例中的传感器模型的生成系统可用于实现本发明实施例中的传感器模型的生成方法。本发明实施例中的传感器测量系统可用于实现本发明实施例中的传感器测量方法。对于本发明系统实施例中未详细披露的细节可参见本发明方法实施例中的相应描述，此处不再一一赘述。
图5为本发明实施例中传感器模型的生成系统的示例性结构图。如图5所示，该系统可包括：第一确定模块501、样本获取模块502、第一模型训练模块503和模型确认模块504。
其中，第一确定模块501用于确定一基线仿真模型的至少一组输入参数样本。
样本获取模块502用于针对每组输入参数样本，将所述输入参数样本输入所述基线仿真模型中，得到对应的一组输出参数，从所述一组输出参数中选取对应预先确定的复数个目标测量位置的输出参数，作为复数个目标传感器的输出参数样本；将所述输入参数样本和所述输出参数样本作为一组样本存储到一样本集中。
第一模型训练模块503用于针对一单个目标传感器或一目标传感器组合，利用所述样本集中第一数量的样本中每个样本的输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对一构建的神经网络模型进行训练，得到对应所述单个目标 传感器或所述目标传感器组合的传感器临时模型；利用所述样本集中剩余的第二数量的样本中每个样本的输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对所述传感器临时模型的输出精度进行校验。
模型确认模块504用于针对至少一个传感器临时模型，在校验精度满足第一设定要求时，将每个传感器临时模型作为所对应的单个目标传感器或目标传感器组合的传感器初始模型。进一步地，模型确认模块504还用于在校验精度不满足第一设定要求时，通知第一确定模块501执行所述确定所述基线仿真模型的至少一组输入参数样本的操作。
图6为本发明实施例中另一种传感器模型的生成系统的示例性结构图。如图6所示，该系统可包括：第二确定模块601、第二模型训练模块602和决策模块603。
第二确定模块601用于将图5所示实施例中训练得到各单个目标传感器和/或目标传感器组合的传感器初始模型的样本集确定为目标样本集。
第二模型训练模块602用于针对按照设定方式确定的单个物理传感器或物理传感器组合需布置的至少一个当前待选目标测量位置，对其他目标测量位置的单个目标传感器或目标传感器组合，从所述目标样本集中每个样本的输出参数样本中选取所述至少一个当前待选目标测量位置对应的至少一个输出参数作为当前辅助输入参数加入所述样本的输入参数样本中，构成当前输入参数样本，利用所述目标样本集中第一数量的样本中每个样本的当前输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对一构建的神经网络模型进行训练，得到对应所述单个目标传感器或所述目标传感器组合的传感器优化模型；利用所述目标样本集中第二数量的样本中每个样本的当前输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对所述传感器优化模型的输出精度进行校验。针对至少一个当前待选目标测量位置，得到由其他目标测量位置的至少一个传感器优化模型构成的一个传感器优化模型集。
决策模块603用于根据各待选目标测量位置下得到的各传感器优化模型集，可根据每个传感器优化模型集的校验精度选取最终的至少一个待选目标测量位置和最终的传感器优化模型集。
具体实现时，图5和图6所示的传感器模型的生成系统可以在不同的设备中实现，或者也可以在同一设备中实现。
图7为本发明实施例中传感器测量系统的示例性结构图。如图7所示，该系统可包括N个传感器模型710和M2个物理传感器720。
其中，N个传感器模型710用于对M个测量位置中的M1个测量位置进行模拟检测。
M2个物理传感器720用于对M个测量位置中剩余的M2个测量位置进行实际检测。
其中，M＝M1+M2，M1大于或等于N；且M、M1和N分别为大于或等于1的整数，M2为大于或等于0的整数。所述N个传感器模型包括N1个传感器初始模型711和N2个以所述M2个物理传感器的输出值为辅助输入参数的传感器优化模型712；其中，N＝N1+N2，N1和N2分别为大于或等于0的整数，且二者不同时为0。
图8为本发明实施例中又一种传感器模型的生成系统的结构示意图，如图8所示，该系统可包括：至少一个存储器81和至少一个处理器82。此外，还可以包括一些其它组件，例如通信端口等。这些组件通过总线83进行通信。
其中：至少一个存储器81用于存储计算机程序。在一个实施方式中，该计算机程序可以理解为包括图5和/或图6所示的传感器模型的生成系统的各个模块。此外，至少一个存储器81还可存储操作系统等。操作系统包括但不限于：Android操作系统、Symbian操作系统、Windows操作系统、Linux操作系统等等。
至少一个处理器82用于调用至少一个存储器81中存储的计算机程序，执行本发明实施例中所述的传感器模型的生成方法。处理器82可以为CPU，处理单元/模块，ASIC，逻辑模块或可编程门阵列等。其可通过所述通信端口进行数据的接收和发送。
需要说明的是，上述各流程和各结构图中不是所有的步骤和模块都是必须的，可以根据实际的需要忽略某些步骤或模块。各步骤的执行顺序不是固定的，可以根据需要进行调整。各模块的划分仅仅是为了便于描述采用的功能上的划分，实际实现时，一个模块可以分由多个模块实现，多个模块的功能也可以由同一个模块实现，这些模块可以位于同一个设备中，也可以位于不同的设备中。
可以理解，上述各实施方式中的硬件模块可以以机械方式或电子方式实现。例如，一个硬件模块可以包括专门设计的永久性电路或逻辑器件(如专用处理器，如FPGA或ASIC)用于完成特定的操作。硬件模块也可以包括由软件临时配置的可编程逻辑器件或电路(如包括通用处理器或其它可编程处理器)用于执行特定操作。至于具体采用机械方式，或是采用专用的永久性电路，或是采用临时配置的电路(如由软件进行配置)来实现硬件模块，可以根据成本和时间上的考虑来决定。
此外，本发明实施例中还提供一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序能够被一处理器执行并实现本发明实施例中所述的传感器模型的生成方法。具体地，可以提供配有存储介质的系统或者装置，在该存储介质上存储着实现上述实施例中任一实施方式的功能的软件程序代码，且使该系统或者装置的计算机(或CPU或MPU)读出并执行存储在存储介质中的程序代码。此外，还可以通过基于程序代码的指令使计算机上操作的操作系统等来完成部分或者全部的实际操作。还可以将从存储介 质读出的程序代码写到插入计算机内的扩展板中所设置的存储器中或者写到与计算机相连接的扩展单元中设置的存储器中，随后基于程序代码的指令使安装在扩展板或者扩展单元上的CPU等来执行部分和全部实际操作，从而实现上述实施方式中任一实施方式的功能。用于提供程序代码的存储介质实施方式包括软盘、硬盘、磁光盘、光盘(如CD-ROM、CD-R、CD-RW、DVD-ROM、DVD-RAM、DVD-RW、DVD+RW)、磁带、非易失性存储卡和ROM。可选择地，可以由通信网络从服务器计算机上下载程序代码。
从上述方案中可以看出，由于本发明实施例中利用基线仿真模型预测值较准确的特点来获取包括大量的由输入参数和对应的输出参数构成的样本的样本集来训练所需物理传感器的神经网络模型，得到对应所需物理传感器的传感器模型(即虚拟传感器)，之后便可利用该传感器模型来代替个别物理传感器，以实现目标设备不适合安装物理传感器的位置的仿真测量；此外通过利用某些传感器模型来代替一些物理传感器，还可以节约成本。
此外，通过利用某些测量位置对应的基线仿真模型输出来作为其它测量位置的辅助输入参数来训练其它测量位置的传感器模型，可进一步提高传感器模型的输出精度。从而在使用中可在所述某些测量位置设置物理传感器，并利用其输出即测量值作为其它测量位置的传感器模型的辅助输入参数，得到其它测量位置的传感器模型的较高精度的输出。
另外，通过将物理传感器、传感器优化模型和传感器初始模型灵活搭配使用，可以应对各种应用场景，提高了目标设备的整体测量精度。
以上所述仅为本发明的较佳实施例而已，并不用以限制本发明，凡在本发明的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本发明的保护范围之内。

Claims (10)

  1. 传感器模型的生成方法，其特征在于，包括：
    确定一基线仿真模型的至少一组输入参数样本(S11)；
    针对每组输入参数样本，将所述输入参数样本输入所述基线仿真模型中，得到对应的一组输出参数，从所述一组输出参数中选取对应预先确定的复数个目标测量位置的输出参数，作为复数个目标传感器的输出参数样本；将所述输入参数样本和所述输出参数样本作为一组样本存储到一样本集中(S12)；
    针对一单个目标传感器或一目标传感器组合，利用所述样本集中第一数量的样本中每个样本的输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对一构建的神经网络模型进行训练，得到对应所述单个目标传感器或所述目标传感器组合的传感器临时模型；利用所述样本集中剩余的第二数量的样本中每个样本的输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对所述传感器临时模型的输出精度进行校验(S13)；
    针对至少一个传感器临时模型，在校验精度满足设定要求时，将每个传感器临时模型作为所对应的单个目标传感器或目标传感器组合的传感器初始模型(S14-S15)。
  2. 根据权利要求1所述的传感器模型的生成方法，其特征在于，进一步包括：在校验精度不满足所述设定要求时，返回执行所述确定至少一组输入参数样本的步骤(S15)。
  3. 根据权利要求1或2所述的传感器模型的生成方法，其特征在于，进一步包括：
    将训练得到各单个目标传感器和/或目标传感器组合的传感器初始模型的样本集作为目标样本集(S21)；
    针对按照设定方式确定的单个物理传感器或物理传感器组合需布置的至少一个当前待选目标测量位置，对其他目标测量位置的单个目标传感器或目标传感器组合，从所述目标样本集中每个样本的输出参数样本中选取所述至少一个当前待选目标测量位置对应的至少一个输出参数作为当前辅助输入参数加入所述样本的输入参数样本中，构成当前输入参数样本，利用所述目标样本集中第一数量的样本中每个样本的当前输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对一构建的神经网络模型进行训练，得到对应所述单个目标传感器或所述目标传感器组合的传感器优化模型；利用所述目标样本集中第二数量的样本中每个样本的当前输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对所述传感器优化模型的输出精度进行校验。针对至少一个当前待选目标测量位置，得到由其他目标测量位置的至少一个传感器优化模型构成的一个传感器优化模型集(S22)；
    根据各待选目标测量位置下得到的各传感器优化模型集，可根据每个传感器优化模型集的校验精度选取最终的至少一个待选目标测量位置和最终的传感器优化模型集(S23)。
  4. 传感器测量方法，其特征在于，包括：
    在M个测量位置中的M1个测量位置利用N个传感器模型进行模拟检测；
    在M个测量位置中剩余的M2个测量位置利用M2个物理传感器进行实际检测；
    其中，M＝M1+M2，M1大于或等于N；且M、M1和N分别为大于或等于1的整数，M2为大于或等于0的整数；
    所述N个传感器包括N1个传感器初始模型和N2个以所述M2个物理传感器的输出值为辅助输入参数的传感器优化模型；其中，N＝N1+N2，N1和N2分别为大于或等于0的整数，且二者不同时为0；
    所述传感器初始模型根据权利要求1或2中所述的传感器模型的生成方法得到；
    所述传感器优化模型和所述M2个物理传感器的测量位置根据权利要求3中所述的传感器模型的生成方法得到。
  5. 传感器模型的生成系统，其特征在于，包括：
    第一确定模块(501)，用于确定一基线仿真模型的至少一组输入参数样本；
    样本获取模块(502)，用于针对每组输入参数样本，将所述输入参数样本输入所述基线仿真模型中，得到对应的一组输出参数，从所述一组输出参数中选取对应预先确定的复数个目标测量位置的输出参数，作为复数个目标传感器的输出参数样本；将所述输入参数样本和所述输出参数样本作为一组样本存储到一样本集中；
    第一模型训练模块(503)，用于针对一单个目标传感器或一目标传感器组合，利用所述样本集中第一数量的样本中每个样本的输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对一构建的神经网络模型进行训练，得到对应所述单个目标传感器或所述目标传感器组合的传感器临时模型；利用所述样本集中剩余的第二数量的样本中每个样本的输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对所述传感器临时模型的输出精度进行校验；
    模型确认模块(504)，用于针对至少一个传感器临时模型，在校验精度满足第一设定要求时，将每个传感器临时模型作为所对应的单个目标传感器或目标传感器组合的传感器初始模型。
  6. 根据权利要求5所述的传感器模型的生成系统，其特征在于，所述模型确认模块(504)进一步用于在校验精度不满足第一设定要求时，通知所述第一确定模块(501) 执行所述确定所述基线仿真模型的至少一组输入参数样本的操作。
  7. 根据权利要求5或6所述的传感器模型的生成系统，其特征在于，进一步包括：
    第二确定模块(601)，用于将训练得到各单个目标传感器和/或目标传感器组合的传感器初始模型的样本集确定为目标样本集；
    第二模型训练模块(602)，用于针对按照设定方式确定的单个物理传感器或物理传感器组合需布置的至少一个当前待选目标测量位置，对其他目标测量位置的单个目标传感器或目标传感器组合，从所述目标样本集中每个样本的输出参数样本中选取所述至少一个当前待选目标测量位置对应的至少一个输出参数作为当前辅助输入参数加入所述样本的输入参数样本中，构成当前输入参数样本，利用所述目标样本集中第一数量的样本中每个样本的当前输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对一构建的神经网络模型进行训练，得到对应所述单个目标传感器或所述目标传感器组合的传感器优化模型；利用所述目标样本集中第二数量的样本中每个样本的当前输入参数样本和对应所述单个目标传感器或所述目标传感器组合的输出参数对所述传感器优化模型的输出精度进行校验。针对至少一个当前待选目标测量位置，得到由其他目标测量位置的至少一个传感器优化模型构成的一个传感器优化模型集；
    决策模块(603)，用于根据各待选目标测量位置下得到的各传感器优化模型集，可根据每个传感器优化模型集的校验精度选取最终的至少一个待选目标测量位置和最终的传感器优化模型集。
  8. 传感器测量系统，其特征在于，包括：
    N个传感器模型(710)，用于对M个测量位置中的M1个测量位置进行模拟检测；和
    M2个物理传感器(720)，用于对M个测量位置中剩余的M2个测量位置进行实际检测；
    其中，M＝M1+M2，M1大于或等于N；且M、M1和N分别为大于或等于1的整数，M2为大于或等于0的整数；
    所述N个传感器模型包括N1个传感器初始模型(711)和N2个以所述M2个物理传感器的输出值为辅助输入参数的传感器优化模型(712)；其中，N＝N1+N2，N1和N2分别为大于或等于0的整数，且二者不同时为0；
    所述传感器初始模型(711)由权利要求5或6中所述的传感器模型的生成系统得到；
    所述传感器优化模型(712)和所述M2个物理传感器的测量位置由权利要求7中所述的传感器模型的生成系统得到。
  9. 传感器模型的生成系统，其特征在于，包括：至少一个存储器(81)和至少一个处理器(82)，其中：
    所述至少一个存储器(81)用于存储计算机程序；
    所述至少一个处理器(82)用于调用所述至少一个存储器(81)中存储的计算机程序，执行如权利要求1至3中任一项所述的传感器模型的生成方法。
  10. 计算机可读存储介质，其上存储有计算机程序；其特征在于，所述计算机程序能够被一处理器执行并实现如权利要求1至3中任一项所述的传感器模型的生成方法。
PCT/CN2019/105197 2019-09-10 2019-09-10 传感器模型的生成方法和系统、传感器测量方法和系统 Ceased WO2021046723A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
EP19945109.7A EP4006767A4 (en) 2019-09-10 2019-09-10 SENSOR PATTERN GENERATION METHOD AND SYSTEM, AND SENSOR MEASUREMENT METHOD AND SYSTEM
PCT/CN2019/105197 WO2021046723A1 (zh) 2019-09-10 2019-09-10 传感器模型的生成方法和系统、传感器测量方法和系统
CN201980096752.3A CN113874866A (zh) 2019-09-10 2019-09-10 传感器模型的生成方法和系统、传感器测量方法和系统

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2019/105197 WO2021046723A1 (zh) 2019-09-10 2019-09-10 传感器模型的生成方法和系统、传感器测量方法和系统

Publications (1)

Publication Number Publication Date
WO2021046723A1 true WO2021046723A1 (zh) 2021-03-18

Family

ID=74866687

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/105197 Ceased WO2021046723A1 (zh) 2019-09-10 2019-09-10 传感器模型的生成方法和系统、传感器测量方法和系统

Country Status (3)

Country Link
EP (1) EP4006767A4 (zh)
CN (1) CN113874866A (zh)
WO (1) WO2021046723A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE102022201637A1 (de) 2022-02-17 2023-08-17 Zf Friedrichshafen Ag Computerimplementiertes Verfahren und System zur Erzeugung eines virtuellen Klons eines Sensors
CN120446390A (zh) * 2025-04-14 2025-08-08 长春大学 气体传感器阵列优化方法、装置、设备和存储介质

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116090282B (zh) * 2022-11-09 2026-03-31 特变电工科技投资有限公司 融合传感器的参数优化方法、装置、电子设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101711390A (zh) * 2007-02-16 2010-05-19 特尔科迪亚许可公司有限公司 用于宽区域监视的有限数量传感器的合理布置的方法
CN109086658A (zh) * 2018-06-08 2018-12-25 中国科学院计算技术研究所 一种基于生成对抗网络的传感器数据生成方法与系统
WO2019012029A1 (en) * 2017-07-12 2019-01-17 Siemens Aktiengesellschaft METHOD AND SYSTEM FOR GAP DETECTION IN SENSOR DATA SETS
CN110007645A (zh) * 2019-04-11 2019-07-12 华中科技大学 一种基于动力学及深度神经网络的进给系统混合建模方法
CN110168570A (zh) * 2016-12-29 2019-08-23 谷歌有限责任公司 用于多个传感器的机器学习虚拟传感器模型

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080154811A1 (en) * 2006-12-21 2008-06-26 Caterpillar Inc. Method and system for verifying virtual sensors
US7787969B2 (en) * 2007-06-15 2010-08-31 Caterpillar Inc Virtual sensor system and method
CN109145815B (zh) * 2018-08-21 2022-05-03 深圳大学 一种sar目标识别方法、装置、计算机设备及存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101711390A (zh) * 2007-02-16 2010-05-19 特尔科迪亚许可公司有限公司 用于宽区域监视的有限数量传感器的合理布置的方法
CN110168570A (zh) * 2016-12-29 2019-08-23 谷歌有限责任公司 用于多个传感器的机器学习虚拟传感器模型
WO2019012029A1 (en) * 2017-07-12 2019-01-17 Siemens Aktiengesellschaft METHOD AND SYSTEM FOR GAP DETECTION IN SENSOR DATA SETS
CN109086658A (zh) * 2018-06-08 2018-12-25 中国科学院计算技术研究所 一种基于生成对抗网络的传感器数据生成方法与系统
CN110007645A (zh) * 2019-04-11 2019-07-12 华中科技大学 一种基于动力学及深度神经网络的进给系统混合建模方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4006767A4 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE102022201637A1 (de) 2022-02-17 2023-08-17 Zf Friedrichshafen Ag Computerimplementiertes Verfahren und System zur Erzeugung eines virtuellen Klons eines Sensors
CN120446390A (zh) * 2025-04-14 2025-08-08 长春大学 气体传感器阵列优化方法、装置、设备和存储介质

Also Published As

Publication number Publication date
EP4006767A4 (en) 2023-05-24
CN113874866A (zh) 2021-12-31
EP4006767A1 (en) 2022-06-01

Similar Documents

Publication Publication Date Title
US8630836B2 (en) Predicting system performance and capacity using software module performance statistics
US20100313070A1 (en) Broken-spheres methodology for improved failure probability analysis in multi-fail regions
US10394640B2 (en) Requirement runtime monitor using temporal logic or a regular expression
KR20190025474A (ko) 플랜트 데이터 예측 장치 및 방법
JP6183449B2 (ja) システム分析装置、及び、システム分析方法
WO2021046723A1 (zh) 传感器模型的生成方法和系统、传感器测量方法和系统
TW201602819A (zh) 調整用於靜態時序分析的時序降額的方法
JP2016045536A (ja) 設計支援装置
JP2015049606A (ja) 管理システム、管理対象装置、管理装置、方法及びプログラム
KR101337216B1 (ko) 컴퓨터 시스템 및 시그니처검증서버
CN115157236A (zh) 机器人刚度模型精度建模方法、系统、介质、设备及终端
CN116244932B (zh) 对车辆进行安全仿真的方法、电子设备及存储介质
CN108885719A (zh) 基于立体视觉的随机地图生成和贝叶斯更新
CN114386159A (zh) 基于数字仿真的航天器机械产品可靠性测试方法
JP6728830B2 (ja) 情報処理装置、情報処理方法、及び、プログラム
US20170140080A1 (en) Performing And Communicating Sheet Metal Simulations Employing A Combination Of Factors
CN111679600B (zh) 控制系统的比较方法、控制终端和计算机可读存储介质
US9245067B2 (en) Probabilistic method and system for testing a material
JP6903246B2 (ja) 情報処理装置、情報処理方法及び情報処理プログラム
CN118534291B (zh) 一种用于测试电源芯片的方法、系统、设备及存储介质
CN115510782B (zh) 定位验证错误的方法、电子设备和存储介质
JP7507997B2 (ja) 保守コスト推定装置及び保守コスト推定方法
US20170184488A1 (en) Facility state analysis device, analysis method for facility state, storage medium, and facility management system
KR101171255B1 (ko) 반도체 집적 회로 설계 시스템 및 방법, 이를 위한 컴퓨터로 읽을 수 있는 기록 매체
WO2024150643A1 (ja) 情報処理装置、情報処理方法、及びコンピュータ読み取り可能な記録媒体

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19945109

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2019945109

Country of ref document: EP

Effective date: 20220228

NENP Non-entry into the national phase

Ref country code: DE