Door body damage detection method, device, electronic device and storage medium

Abstract

The door body damage detection method, device, electronic device and storage medium provided by the present invention belong to the field of artificial intelligence technology, and include: obtaining a door body damage recognition agent through the reinforcement learning of a door body damage detection model; using the door body damage recognition agent to recognize the opening angle sequence and opening force sequence of a to-be-detected door body within a preset time sequence, and outputting a damage coefficient sequence of each door body core component of the to-be-detected door body within the preset time sequence; and determining the detection result of the to-be-detected door body based on the damage coefficient sequence of each door body core component. After establishing a 3D parametric damage behavior simulation model considering damage, the present invention reduces the order of the discrete damage behavior simulation model through deep learning to obtain a door body damage detection model, and then learns the door body damage detection model considering damage evolution behavior based on a reinforcement learning algorithm to obtain a door body damage recognition agent, which can realize the real-time monitoring of the health state of the door body system.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular, to a method, device, electronic device, and storage medium for detecting door body damage. Background Art

[0002] In the long-term use of the door body system of electrical appliances such as dishwashers, core component damage may occur, affecting the performance and safety of the device. Therefore, it is a current industry requirement to monitor the damage status of the door body system in real time and accurately to improve the reliability and service life of the device.

[0003] In the traditional health management mode of the door body system, generally there are two ways to obtain the health status of the door body system: 1) Collect the usage feedback of customers. During the usage process of customers, if they find that the door body system has abnormalities, they will feedback to the customer service, and the after-sales team will check the abnormal situation to find the cause of the failure of the door body system. Through long-term accumulation, relevant failure data can be collected. 2) Establish a corresponding laboratory to conduct the quick closing and accelerating fatigue durability life test of the door body system. By using this method, the evolution of the health status of the door body system can be understood, but due to the difference in the action timing of the door body system in the actual usage process, the result of damage evolution is inaccurate.

[0004] It can be seen that there is no technology for real-time monitoring of the damage of the door body system in the current industry, and it is urgent to fill this industry gap. Summary of the Invention

[0005] The present invention provides a method, device, electronic device, and storage medium for detecting door body damage, so as to solve the defect that the current detection of the door body system of household appliances cannot achieve real-time and accurate detection.

[0006] The present invention provides a method for detecting door body damage, including the following steps:

[0007] Through the reinforcement learning of the door body damage detection model, obtain the door body damage recognition agent;

[0008] Use the door body damage recognition agent to identify the opening angle sequence and opening force sequence of the door body to be tested within a preset time sequence, and output the damage coefficient sequence of each core component of the door body to be tested within the preset time sequence;

[0009] Based on the damage coefficient sequence of each core component of the door body, determine the detection result of the door body to be tested.

[0010] According to the method for detecting door body damage provided by the present invention, the door body damage detection model is constructed by the following steps:

[0011] Construct a damage behavior simulation model for each discrete time point based on the physical benchmark data and the damage coefficients of each core component of the door body at each discrete time point in the preset time sequence; the physical benchmark data is obtained by conducting physical opening and closing experiments on each core component of the real physical door body.

[0012] Train the first initial network model using the training data set to obtain the door body damage detection model; any training data in the training data set includes the opening angle of a damage behavior simulation model at its corresponding discrete time point and the damage coefficients of each core component of the door body, and the label of the training data is the opening force at the discrete time point.

[0013] According to a door body damage detection method provided by the present invention, the constructing a damage behavior simulation model for each discrete time point based on the physical benchmark data and the damage coefficients of each core component of the door body at each discrete time point in the preset time sequence includes:

[0014] Use the physical benchmark data to benchmark each undamaged simulation model one by one to obtain a normal behavior simulation model for each core component of the door body.

[0015] Combine the normal behavior simulation models to form a door body system simulation model.

[0016] Update the door body system simulation model according to the damage coefficients of each core component of the door body at each discrete time point in the preset time sequence to obtain the damage behavior simulation models in different damage states at each discrete time point.

[0017] According to a door body damage detection method provided by the present invention, the physical benchmark data includes force time sequence curve samples obtained by conducting physical opening and closing experiments under angle control for each core component of the door body respectively.

[0018] The using the physical benchmark data to benchmark each undamaged simulation model one by one to obtain a normal behavior simulation model for each core component of the door body includes:

[0019] For the undamaged simulation model of any core component of the door body, adjust the model parameters of the undamaged simulation model according to the curve difference between the virtual force time sequence curve output by the undamaged simulation model and the force time sequence curve sample until the curve difference is within the preset range, and obtain the adjusted undamaged simulation model as the normal behavior simulation model.

[0020] A door body damage detection method provided by the present invention, wherein the core components of the door body include a complete door body system and all key subsystems, and the key subsystems include one or more of a hinge subsystem, a drawstring spring subsystem, a lock subsystem, and a sealing subsystem; the hinge subsystem is composed of a skeleton subsystem and an outer door subsystem;

[0021] Correspondingly, the non-destructive simulation model includes a complete door body system simulation model, a hinge subsystem simulation model, a drawstring spring subsystem simulation model, a lock subsystem simulation model, and a sealing subsystem simulation model.

[0022] A door body damage detection method provided by the present invention, wherein the normal behavior simulation models are combined to form a door body system simulation model, including:

[0023] Fuse the normal behavior simulation models of the complete door body system, the hinge subsystem, the drawstring spring subsystem, the lock subsystem, and the sealing subsystem corresponding to the complete door body system simulation model, the hinge subsystem simulation model, the drawstring spring subsystem simulation model, the lock subsystem simulation model, and the sealing subsystem simulation model to obtain the door body system simulation model.

[0024] A door body damage detection method provided by the present invention, wherein the door body system simulation model is updated according to the damage coefficient of each door body core component at each discrete time point in the preset time sequence, and the damage behavior simulation model in different damage states at each discrete time point is obtained, including:

[0025] Based on the current opening and closing times of any door body core component at any discrete time point and the fatigue opening and closing times limit value of the any door body core component, determine the damage coefficient of the any door body core component at the any discrete time point;

[0026] Based on the damage coefficient, determine the scaling coefficient of the target parameter related to the door body damage of the any door body core component;

[0027] At the any discrete time point, update the door body system simulation model according to the scaling coefficients of the target parameters related to the door body damage of all door body core components, and obtain the damage behavior simulation model at the any discrete time point;

[0028] Traverse all the discrete time points to obtain all the damage behavior simulation models.

[0029] A door body damage detection method provided by the present invention, the target parameters of the hinge subsystem include the deformation parameters at the connection point between the hinge and the pulling rope, the target parameters of the pulling rope spring subsystem include the pulling rope stiffness and the telescopic spring stiffness, the target parameters of the lock subsystem include the locking spring stiffness, and the target parameters of the sealing subsystem include the resilience of the sealing strip.

[0030] A door body damage detection method provided by the present invention, using any training data in the training dataset to train the first initial network model to obtain the door body damage detection model, including:

[0031] According to the preset time sequence, sequentially input the training data corresponding to each discrete time point into the first initial network model to obtain an opening force prediction sequence output by the first initial network model;

[0032] According to the network error between the opening force prediction sequence and the opening force label sequence, gradually update the model parameters of the first initial network model until the model converges to obtain the door body damage detection model;

[0033] The opening force label sequence is composed of the opening forces at all discrete time points in the preset time sequence.

[0034] A door body damage detection method provided by the present invention, the obtaining of the door body damage recognition agent through the reinforcement learning of the door body damage detection model includes:

[0035] Using the door body damage detection model to construct a reinforcement training set; any reinforcement training data in the reinforcement training set includes the opening angle and the opening force at any discrete time point, and the label of any reinforcement training data is the damage coefficient of each door body core component at any discrete time point, and the opening force is obtained after inputting the opening angle and the damage coefficient of each door body core component into the door body damage detection model;

[0036] Using the reinforcement training set to train the second initial network model to obtain the door body damage recognition agent.

[0037] A door body damage detection method provided by the present invention, the using the reinforcement training set to train the second initial network model to obtain the door body damage recognition agent includes:

[0038] According to the preset time sequence, sequentially input the reinforcement training data corresponding to each discrete time point into the second initial network model to obtain a damage coefficient prediction sequence output by the second initial network model;

[0039] Predict the network error between the damage coefficient prediction sequence and the damage coefficient label sequence, and gradually update the model parameters of the second initial network model until the model converges to obtain the door body damage recognition agent;

[0040] The damage coefficient label sequence is composed of the damage coefficients of all core components of the door body at all discrete time points in the preset time series.

[0041] According to a door body damage detection method provided by the present invention, both the first initial network model and the second initial network model are obtained based on the long short-term memory network model as the basic model architecture.

[0042] The present invention also provides a door body damage detection device, including:

[0043] An agent training unit for obtaining a door body damage recognition agent through reinforcement learning of the door body damage detection model;

[0044] An agent recognition unit for using the door body damage recognition agent to recognize the opening angle sequence and opening force sequence of the door body to be tested within a preset time series, and outputting the damage coefficient sequence of all core components of the door body to be tested within the preset time series;

[0045] A door body detection unit for determining the detection result of the door body to be tested based on the damage coefficient sequence of all core components of the door body.

[0046] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the door body damage detection method as described in any one of the above.

[0047] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the door body damage detection method as described in any one of the above.

[0048] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the door body damage detection method as described in any one of the above.

[0049] The door body damage detection method, device, electronic device, and storage medium provided by the present invention can realize real-time monitoring of the health status of the door body system after establishing a 3D parametric damage behavior simulation model considering damage, reducing the order of the discrete damage behavior simulation model through deep learning to obtain a door body damage detection model, and then learning the door body damage detection model considering damage evolution behavior based on the reinforcement learning algorithm to obtain a door body damage recognition agent. Description of the Drawings

[0050] To more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0051] Figure 1 It is one of the schematic flowcharts of the door body damage detection method provided by the present invention.

[0052] Figure 2 It is the second schematic flowchart of the door body damage detection method provided by the present invention.

[0053] Figure 3 It is the schematic flowchart of the construction method of the door body damage detection model provided by the present invention.

[0054] Figure 4 It is the schematic diagram of the method for determining the damage coefficient of the hinge subsystem provided by the present invention.

[0055] Figure 5 It is the schematic diagram of the sealing strip stiffness curve provided by the present invention.

[0056] Figure 6 It is the schematic flowchart of the training process of the door body damage detection model provided by the present invention.

[0057] Figure 7 It is the schematic structural diagram of the door body damage detection device provided by the present invention.

[0058] Figure 8 It is the first schematic diagram of the performance of the door body damage detection model provided by the present invention in the training dataset.

[0059] Figure 9 It is the second schematic diagram of the performance of the door body damage detection model provided by the present invention in the training dataset.

[0060] Figure 10 It is the third schematic diagram of the performance of the door body damage detection model provided by the present invention in the training dataset.

[0061] Figure 11 It is the fourth schematic diagram of the performance of the door body damage detection model provided by the present invention in the training dataset.

[0062] Figure 12 It is the fifth schematic diagram of the performance of the door body damage detection model provided by the present invention in the training dataset.

[0063] Figure 13 It is the first schematic diagram of the performance of the door body damage detection model provided by the present invention in the test dataset.

[0064] Figure 14 It is the second schematic diagram of the performance of the door body damage detection model provided by the present invention on the test data set.

[0065] Figure 15 It is the third schematic diagram of the performance of the door body damage detection model provided by the present invention on the test data set.

[0066] Figure 16 It is the fourth schematic diagram of the performance of the door body damage detection model provided by the present invention on the test data set.

[0067] Figure 17 It is the fifth schematic diagram of the performance of the door body damage detection model provided by the present invention on the test data set.

[0068] Figure 18 It is one of the schematic diagrams of the evolution of the loss function value during the training process of the door body damage detection model provided by the present invention.

[0069] Figure 19 It is the second schematic diagram of the evolution of the loss function value during the training process of the door body damage detection model provided by the present invention.

[0070] Figure 20 It is the schematic diagram of the structure of the electronic device provided by the present invention. Specific embodiments

[0071] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts shall fall within the protection scope of the present invention.

[0072] It should be noted that in the description of the present invention, the terms "include", "comprise" or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the existence of additional identical elements in the process, method, article or device comprising the said element. The orientation or positional relationship indicated by terms such as "upper", "lower", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation to the present invention. Unless otherwise clearly specified and defined, the terms "mounted", "connected" and "coupled" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0073] The terms "first", "second", etc. in the present invention are used to distinguish similar objects, rather than to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances, so that the embodiments of the present invention can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally of the same kind, and do not limit the number of objects. For example, the first object may be one or multiple.

[0074] For the scenario of health status management of electrical appliances, especially household electrical appliances, it is a very broad category. Taking a dishwasher as an example, during the actual use by users, the scenarios that require health status management are as follows: 1) Monitoring the health status of the dishwasher's washing system: During the washing process, whether there are abnormalities in water pressure, water temperature, booster pump, breather, drainage system, etc. The monitoring, prediction and fault diagnosis of the overall health status of the washing system are generally related to the relevant control system; 2) Monitoring the health status of the dishwasher's structural system: Whether there are damages to the various systems of the dishwasher, including the door system, inner tank system, chassis system, etc., analyzing the damage value, remaining life, and judging whether the life limit has been reached, etc.

[0075] During the user's operation, for the door system of electrical equipment, such as the core components of the door body like the pull cord spring subsystem, the sealing strip subsystem, the lock subsystem, and the hinge subsystem, after a certain number of years of use, there is a probability of failure. For example, the following failure situations may occur: 1) Pull cord spring subsystem: The telescopic spring becomes loose or breaks, and the pull cord wears or breaks; 2) Sealing strip subsystem: The sealing performance of the sealing strip gradually decreases, and the resilience decreases; 3) Lock subsystem: The insertion position of the lock bolt wears, and the locking spring becomes loose or breaks; 4) Hinge subsystem: The hinge structure is deformed or broken, etc.

[0076] During the design process of the door system, in order to quickly obtain the damage evolution process of the entire life cycle of the door system, a corresponding laboratory will be established to conduct the rapid closing acceleration fatigue durability life test of the door system. At the same time, combined with the failure data obtained from the failure analysis based on the feedback during the customer's use process, however, since the action timing of each door system is complex and personalized, resulting in different speeds of damage evolution, the existing these methods often cannot achieve systematic and real-time analysis of the damage of the door system. How to solve this need is an urgent problem in the industry.

[0077] Today, with the rapid development of digital technology, the present invention provides a door body damage detection method, device, electronic device, and storage medium based on the integration of digital twin technology and reinforcement learning algorithms and deep learning algorithms, in order to provide a new health management mode for the door system.

[0078] The present invention proposes a door body damage detection method, device, electronic device, and storage medium, and its main implementation steps include the following steps:

[0079] First, design a model experiment to obtain physical benchmark data as the benchmark standard for the non-destructive simulation model. Secondly, construct a normal behavior simulation model through the benchmark of the physical benchmark data. Thirdly, based on the corrected normal behavior simulation model, using the fatigue damage theory, put forward reasonable mechanical assumptions, establish a damage behavior simulation model considering damage evolution, and output a training data set for deep learning. Then, train based on the deep learning network algorithm, and output the corresponding AI reduced-order model of the damage behavior simulation model as a high-fidelity digital twin model considering damage evolution behavior, that is, the door body damage detection model. Subsequently, based on the reinforcement learning algorithm, the door body damage detection model can be learned to obtain the corresponding intelligent agent network model, which can be used for real-time monitoring of the damage state of the door system.

[0080] The following combines Figures 1 - 20 Describe the door body damage detection method, device, electronic device, and storage medium provided by the present invention.

[0081] Figure 1It is one of the schematic flowcharts of the door body damage detection method provided by the present invention. As Figure 1 shown, it includes but is not limited to the following steps:

[0082] Step 101, through the reinforcement learning of the door body damage detection model, obtain the door body damage recognition agent.

[0083] First, a high-fidelity door body damage detection model needs to be constructed. This door body damage detection model is constructed based on digital twin technology and deep learning algorithms, and can simulate the behavior of the door body system in different damage states. The specific construction steps mainly include but are not limited to:

[0084] Conduct physical opening and closing tests on the core components of the real physical door body (such as hinge subsystem, pull rope spring subsystem, lock button subsystem, and sealing subsystem) to obtain physical benchmark data. In the test, the angle control method can be used to record the opening force data of the door body at different opening and closing angles, and form a force time series curve sample.

[0085] Based on the physical benchmark data, benchmark each non-destructive simulation model to obtain the normal behavior simulation model of each door body core component. Combine these normal behavior simulation models to form a complete door body system simulation model.

[0086] According to the fatigue characteristics of each door body core component, define the damage coefficient, and update the relevant parameters in the door body system simulation model based on the damage coefficient to generate the damage behavior simulation models in different damage states.

[0087] Use the training data set to train the initial network model (such as long short-term memory network LSTM) to obtain the door body damage detection model. The training data includes the opening angle, damage coefficient of the damage behavior simulation model at each discrete time point, and the corresponding opening force label.

[0088] Finally, optimize the door body damage detection model through the reinforcement learning algorithm to obtain the door body damage recognition agent. The specific implementation steps can be described as follows:

[0089] Define the behavior simulation environment of the door body system as the environment of reinforcement learning, where the state is the sequence of opening angle and damage coefficient, and the action is the decision variable in the space (such as the change of opening force).

[0090] Design the reward function according to the closeness between the opening force output by the model and the real damage state. For example, when the error between the output opening force sequence and the actual opening force sequence is small, give a positive reward; otherwise, give a negative reward.

[0091] Train the door body damage detection model using reinforcement learning algorithms (such as Q-learning, DQN, or PPO) so that the agent can learn the optimal policy in a given environment and output an accurate sequence of damage coefficients.

[0092] Step 102: Use the door body damage recognition agent to recognize the opening angle sequence and opening force sequence of the door body to be tested within a preset time sequence, and output the damage coefficient sequence of each core component of the door body to be tested within the preset time sequence.

[0093] In practical applications, when detecting the damage of a door body to be tested, it is necessary to collect its opening angle sequence and opening force sequence within a preset time sequence. The specific method is as follows:

[0094] Install high-precision force sensors and angle sensors on the door body to be tested, and collect opening force and opening angle data in real time.

[0095] Adopt an angle control method to ensure that the angle increment of each door opening is consistent (for example, increasing by 1° each time), and record the opening force and opening angle at each discrete time point.

[0096] Input the collected opening angle sequence and opening force sequence into the door body damage recognition agent trained by reinforcement learning. The door body damage recognition agent will output the damage coefficient sequence of each core component of the door body to be tested within the preset time sequence according to the learned policy.

[0097] Step 103: Determine the detection result of the door body to be tested based on the damage coefficient sequence of each core component of the door body.

[0098] The damage degree can be evaluated according to the damage coefficient sequence of each core component of the door body. For example: when the damage coefficient of a certain core component of the door body is close to 0, it means that the core component of the door body is in good condition; when the damage coefficient of a certain core component of the door body is close to 1, it means that the core component of the door body is close to the service life limit.

[0099] Finally, the overall health status of the door body to be tested can be determined by synthesizing the damage coefficient sequences of each core component of the door body. The specific method is as follows: If the damage coefficients of at least 1 core component of the door body are all high, it is determined that the overall damage of the door body to be tested is serious and maintenance or replacement is required; if the damage coefficients of most core components are low, it is determined that the door body to be tested is in good condition and can continue to be used.

[0100] Figure 2 This is the second flow diagram of the door body damage detection method provided by the present invention. The following will be described by taking the detection of the dishwasher door body system as an example Figure 2 As shown, taking the detection of the dishwasher door body system as an example for illustration.

[0101] Taking the dishwasher door system as an example, the core components of its door body include a hinge subsystem, a pull cord spring subsystem, a latch subsystem, a sealing subsystem, etc. Through the above method, real-time damage detection of the dishwasher door system can be achieved, and the specific steps are as follows:

[0102] First, construct normal behavior simulation models for each core component of the dishwasher door system, and combine all normal behavior simulation models to form a door system simulation model. Then, introduce a damage evolution mechanism to generate damage behavior simulation models under different damage states, so as to construct a training data set for model training of the initial network model by using the damage behavior simulation models.

[0103] Sequentially input the training data corresponding to each discrete time point in the training data set into the initial network model, including the input data and label data recorded at each discrete time point in the training data set, and input the input data (opening angle and damage coefficient) among them into the initial network model.

[0104] Run the initial network model to obtain the predicted opening force sequence output by it, that is, use the initial network model to infer the input data and generate a predicted opening force sequence (collect the prediction results at each discrete time point to form a complete predicted opening force sequence).

[0105] Use the mean square error (MSE) or other statistical metrics to calculate the network error between the predicted sequence and the label sequence. According to the calculated network error, gradually update the parameters of the initial network model until the model converges to obtain a door body damage detection model. This door body damage detection model can accurately predict the corresponding opening force based on the input opening angle and damage coefficient.

[0106] Finally, construct a reinforcement training set through the door body damage detection model (use the opening angle sequence and opening force sequence simulated by the door body damage detection model as input, and the corresponding damage coefficient as the label to construct a reinforcement learning training set), so as to use the generated reinforcement learning training set to train a new initial network model to obtain the required door body damage recognition agent.

[0107] Analyze the real-time collected opening force and opening angle data by using the door body damage recognition agent, and the damage coefficient sequence of each core component of the door body within the preset time sequence can be obtained. Based on the damage coefficient sequence of each core component of the door body, the damage degree of each core component of the door body can be evaluated. By synthesizing the damage coefficient sequences of each core component of the door body, the overall health state of the door body to be tested can be determined quickly and accurately.

[0108] The door body damage detection method provided by the present invention, after establishing a 3D parametric damage behavior simulation model considering damage, reduces the order of the discrete damage behavior simulation model through deep learning to obtain a door body damage detection model, and then learns the door body damage detection model considering damage evolution behavior based on a reinforcement learning algorithm to obtain a door body damage recognition agent, which can realize real-time monitoring of the health state of the door body system.

[0109] The following details the construction method of the door body damage detection model provided by the present invention.

[0110] Figure 3 It is a schematic flow chart of the construction method of the door body damage detection model provided by the present invention, as Figure 3 shown, including but not limited to the following steps:

[0111] Step 301, construct a damage behavior simulation model for each of the discrete time points based on physical calibration data and damage coefficients of each core component of the door body at each discrete time point in a preset time sequence.

[0112] Among them, the physical calibration data is obtained by conducting physical opening and closing experiments on each core component of a real physical door body. The specific acquisition method can be:

[0113] Select a representative dishwasher door body system to ensure that each of its core components (such as the pull rope spring subsystem, the sealing subsystem, the lock buckle subsystem, etc.) is in an initial undamaged state. In the physical opening and closing experiment, an angle control method is adopted, that is, the angle increment of each door opening is kept consistent (for example, increased by 1° each time), and the opening force data of the door body system at different opening and closing angles is recorded. Specifically, high-precision force sensors and angle sensors can be used to collect the opening force data and the opening angle at each discrete time point in the preset time sequence (i.e., discrete sampling time points), constituting the opening force time sequence data and the opening angle time sequence data. The obtained opening force time sequence data and opening angle time sequence data can be represented by an opening force time sequence curve and an opening angle time sequence curve respectively. These data will be used as physical calibration data for subsequent calibration of the undamaged simulation model.

[0114] The damage coefficient is mainly used to quantify the damage degree of each core component of the door body at different usage stages, and generally can be determined by the ratio between the current opening and closing times of each core component of the door body and the fatigue opening and closing times limit value.

[0115] For example, for the pull rope spring subsystem, its damage coefficient D at any discrete time point can be defined as: , where is the current opening and closing times at this any discrete time point, is the fatigue opening and closing times limit value of the pull rope spring subsystem.

[0116] Further, based on the physical benchmark data and damage coefficients, a damage behavior simulation model for each discrete time point can be constructed, and the following steps can be used to achieve this:

[0117] First, according to the initial parameters of each core component of the door body, an initialization simulation model of each core component of the door body system is constructed. This initialization simulation model can be called a non-damage simulation model.

[0118] Then, the physical benchmark data is used to benchmark each non-damage simulation model respectively, and a normal behavior simulation model of each core component of the door body can be obtained. Using the normal behavior simulation models of all core components of the door body, a door body system simulation model can be constructed.

[0119] Finally, at each discrete time point in the preset time sequence, according to the loss coefficient of each core component of the door body at this discrete time point, the parameters of the corresponding components in the door body system simulation model are updated. For example, at a certain discrete time point, the stiffness of the pull rope spring is adjusted according to the damage coefficient D of the pull rope spring subsystem. In this way, at each discrete time point, a damage behavior simulation model considering the damage evolution behavior can be obtained.

[0120] Assume that 50 discrete sampling points are set in the entire preset time sequence, then 50 damage behavior simulation models in different damage states can be obtained.

[0121] Step 302: Use the training data set to train the initial network model to obtain the door body damage detection model.

[0122] Among them, any training data in the training data set includes the opening angle of a damage behavior simulation model at its corresponding discrete time point and the damage coefficients of each core component of the door body, and the label of the training data is the opening force at the discrete time point.

[0123] Specifically, the preparation of the training data set can be carried out first. Each training data in this training data set is closely related to the damage behavior simulation model obtained at each discrete sampling point. First, through experiments, the opening force data of the door body system at different opening and closing angles within the preset time sequence is tested first, that is, the opening angle and opening force at each discrete sampling point can be obtained. At the same time, according to the number of opening and closing times of each core component of the door body system at each discrete sampling point, the damage coefficients of each core component of the door body are counted.

[0124] Further, the opening angle obtained at each discrete time point and the damage coefficients of each core component of the door body are combined into a training data, and the opening force obtained at this discrete time point is used as the label of this training data, and a set of training samples corresponding to each discrete time point can be obtained.

[0125] Suppose there are 5 core components of the door body, and the damage coefficients at any discrete time point are D1, D2, D3, D4, and D5 respectively. Then the input of the training sample at this discrete time point = (door opening angle, D1, D2, D3, D4, D5), and its corresponding label = (door opening force).

[0126] Select a suitable neural network model as the first initial network model. For example, after selecting the long short-term memory network (LSTM), use the training data set to train it. The specific training steps can be as follows:

[0127] (1) Input a training data in the training data set into the first initial network model.

[0128] (2) Calculate the predicted value of the door opening force output by the first initial network model, and calculate the network error (such as the mean square error) between the predicted value of the door opening force and the label corresponding to this training data.

[0129] (3) Use the backpropagation algorithm to update the parameters of the first initial network model to optimize the model performance. The goal is to reduce the network error between the output of the first initial network model and the label.

[0130] (4) Repeat the above process until the model converges to obtain the trained door body damage detection model.

[0131] The method for constructing the door body damage detection model provided by the present invention calibrates the simulation model through a designed model experiment, and at the same time establishes a 3D parametric damage behavior simulation model considering damage. Finally, through deep learning, the discrete damage behavior simulation model is reduced in order to obtain a network model that can quickly and accurately monitor the door body damage in real time, filling the gap in the industry.

[0132] Based on the content of the above embodiments, as an alternative embodiment, the core components of the door body include a complete door body system and all key subsystems. The key subsystems include one or more of a hinge subsystem, a cable spring subsystem, a lock subsystem, and a sealing subsystem; the hinge subsystem is composed of a skeleton subsystem and an outer door subsystem;

[0133] Correspondingly, the non-destructive simulation model includes a complete door body system simulation model, a hinge subsystem simulation model, a cable spring subsystem simulation model, a lock subsystem simulation model, and a sealing subsystem simulation model.

[0134] Taking the door system of a dishwasher as an example, its key subsystems mainly include a hinge subsystem, a drawstring spring subsystem, a lock subsystem, and a sealing subsystem. Overall, the above key subsystems are composed of 6 core components: springs (to distinguish, the spring of the drawstring spring subsystem is called a telescopic spring, and the spring of the lock subsystem is called a locking spring), ropes, pulleys, hinge walls, the door body, door lock buckles, and sealing strips.

[0135] Correspondingly, the simulation model of the entire dishwasher door system is formed by fusing the simulation models of 5 subsystems related to the core components of the door body, namely the complete door system simulation model, the simulation model of the hinge subsystem, the simulation model of the drawstring spring subsystem, the simulation model of the lock subsystem, and the simulation model of the sealing subsystem.

[0136] Traditional benchmarking methods require obtaining the key mechanical parameters of each core component of the door body through standard mechanical test experiments. For standard mechanical test experiments, it is generally necessary to fabricate standard specimens for each core component of the door body, and it may be necessary to complete the acquisition of key mechanical parameters through multiple test experiments. Many universities or research institutes are unable to complete all the test experiments, and generally, the test cycle is relatively long and the test cost is relatively high.

[0137] In view of this, the present invention benchmarks the simulation model of the dishwasher door system by designing key model experiments and directly benchmarks at the system level.

[0138] First, the following will introduce how to construct the benchmark models of each real core component of the door body.

[0139] The hinge subsystem mainly includes a skeleton and an outer door. To be consistent with the non-destructive simulation model to be constructed later, the state of its benchmark model experiment is designed as: removing the lock, sealing strip, and rope, and retaining the door hinge.

[0140] The drawstring spring subsystem mainly includes a skeleton, an outer door, a drawstring, a spring, etc. To be consistent with the non-destructive simulation model to be constructed later, the state of its benchmark model experiment is designed as: removing the lock, sealing strip, and retaining the rope, outer door, hinge, etc.

[0141] The lock subsystem mainly includes a skeleton, an outer door, and a lock. To be consistent with the non-destructive simulation model to be constructed later, the state of its benchmark model experiment is designed as: removing the sealing strip, rope, etc., and retaining the lock, outer door, hinge, etc.

[0142] The sealing subsystem mainly includes a skeleton, an outer door, and a sealing strip, etc. To be consistent with the non-destructive simulation model to be constructed later, the state of its benchmark model experiment is designed as: removing the lock, rope, and retaining the outer door, hinge, and sealing strip, etc.

[0143] Taking the door system of the above dishwasher as an example, a damage behavior simulation model for each discrete time point in the preset time sequence is constructed by using the physical benchmark data and the damage coefficients of each core door component, specifically including but not limited to:

[0144] Using the physical benchmark data to benchmark each non-damage simulation model one by one to obtain the normal behavior simulation model of each core door component;

[0145] Combining the normal behavior simulation models to form a door system simulation model;

[0146] According to the damage coefficients of each core door component at each discrete time point in the preset time sequence, update the door system simulation model to obtain the damage behavior simulation models under different damage states at each discrete time point.

[0147] The non-damage simulation model is established based on the theoretical and empirical data of each core door component and is used to simulate the behavior of the core door component in a non-damaged state. The general construction process includes:

[0148] 1) Construct non-damage simulation models according to the physical characteristics and mechanical behaviors of each core door component. These non-damage simulation models mainly include: hinge subsystem, pull rope spring subsystem, lock subsystem, sealing subsystem, etc.

[0149] 2) Initialize the model parameters related to each core door component in these non-damage simulation models, such as the stiffness, strength, friction coefficient, etc. of each component.

[0150] Furthermore, use the physical benchmark data to benchmark the above non-damage simulation models to adjust the model parameters of each non-damage simulation model so that the simulation output of the non-damage simulation model is consistent with the real test data, and obtain the normal behavior simulation models of each core door component.

[0151] As an optional embodiment, to simplify and standardize the benchmark process, when using the physical benchmark data to benchmark the above non-damage simulation models, the following settings can be made for the corresponding benchmark process:

[0152] (1) During the benchmark test process of the non-damage simulation model of each core door component, since the door experiences the same time sequence actions, from the door locked state to opening to a preset angle and then returning to the original locked state, collect relevant data, including recording the magnitude of each opening force and the corresponding opening angle at discrete time points.

[0153] (2) The test process uses angle control, that is, in each opening and closing process of each test, the time used for opening 1° is the same. Through the above tests, the force time series curve and angle time series curve of the real physical prototype can be obtained, which are used as the physical benchmark data for subsequent simulation benchmarking.

[0154] (3) A force sensor is used to collect the magnitude of the opening force, and an angle sensor is used to collect the opening angle. The entire opening process can be automatically controlled by a robotic arm.

[0155] Further, the normal behavior simulation models of each core component of the door body are combined to form a complete door body system simulation model. Specifically, the normal behavior simulation models of each core component of the door body, such as the cable spring subsystem, seal strip subsystem, lock subsystem, hinge subsystem, etc., are integrated together to form a complete door body system simulation model. System-level verification can also be performed on the integrated door body system simulation model to ensure that the interaction and overall behavior between each core component of the door body conform to the actual physical system.

[0156] The damage coefficient is used to quantify the damage degree of each core component of the door body at different usage stages. First, according to the fatigue characteristics of the core components of the door body, the damage coefficient D of each core component of the door body can be defined. i The value range of each damage coefficient is [0, 1], where 0 indicates no damage and 1 indicates complete failure.

[0157] Then, according to the current opening and closing times and the fatigue opening and closing times limit of each core component of the door body, the damage coefficient at each discrete time point is calculated.

[0158] Further, according to the determined damage coefficient of each core component of the door body at each discrete time point, the relevant model parameters in the door body system simulation model can be updated. For example, if the damage coefficient of the cable spring subsystem is calculated as D2 at a certain discrete time point, the stiffness of the cable spring in the cable spring subsystem of the door body system simulation model can be updated according to the damage coefficient D2.

[0159] In this way, by continuously running within the preset time series and obtaining the updated door body system simulation model, the damage behavior simulation model at each discrete time point is obtained.

[0160] The method for constructing the door body damage detection model provided by the present invention obtains the normal behavior simulation models of each component by using physical calibration data to calibrate the non-destructive simulation models of each core component of the door body. Subsequently, these normal behavior simulation models are combined to form a complete door body system simulation model. Finally, according to the damage coefficients at each discrete time point, the door body system simulation model is updated to obtain the damage behavior simulation models in different damage states. This process solves the technical problem that the simulation software cannot directly establish damage evolution, provides a solid foundation for the subsequent construction of the door body damage detection model, and ensures the accuracy and reliability of the model.

[0161] As an optional embodiment, the physical calibration data includes force time series curve samples obtained by performing physical opening and closing experiments on each of the core components of the door body under angle control.

[0162] The following details how to calibrate each non-destructive simulation model one by one using the physical calibration data to obtain the normal behavior simulation models of each core component of the door body. The specific process includes:

[0163] For the non-destructive simulation model of any core component of the door body, according to the curve difference between the virtual force time series curve output by the non-destructive simulation model and the force time series curve sample, adjust the model parameters of the non-destructive simulation model until the curve difference is within a preset range, and obtain the adjusted non-destructive simulation model as the normal behavior simulation model.

[0164] The physical calibration data is obtained by performing physical opening and closing experiments on each core component of the real physical door body. Taking the dishwasher door body system as an example, its actual acquisition process can be realized through the following steps:

[0165] First, select a representative dishwasher door body system to ensure that all its important components (such as drawstring springs, sealing strips, lock catches, etc.) are in an initial non-damaged state.

[0166] In the experiment, adopt the angle control method, that is, keep the angle increment of each door opening consistent (for example, increase by 1° each time), and record the door opening force data at different opening and closing angles of the door body.

[0167] At the same time, use high-precision force sensors and angle sensors to collect the time series data of the door opening force and the door opening angle respectively to form force time series curve samples, and these physical calibration data will be used as the basis for subsequent simulation model calibration.

[0168] The lossless simulation model is pre-constructed using simulation software (such as MotionView) based on the physical characteristics and mechanical behaviors of the core components of each door body. For the dishwasher door body system, it mainly includes the hinge subsystem simulation model, the cable spring subsystem simulation model, the lock subsystem simulation model, the sealing subsystem simulation model, etc. The hinge subsystem simulation model is mainly composed of the skeleton subsystem simulation model and the outer door subsystem simulation model.

[0169] The skeleton subsystem simulation model is a simulation model mainly composed of components such as the inner tank, the inner tank hinge, and the door hinge. MotionView and MotionSolve are used as the front and back processors respectively. Unit selection is carried out according to the force characteristics of each component. All components of the inner tank, the inner tank hinge, and the door hinge are simulated using rigid body elements; all components of the inner tank are fixedly connected, the inner tank and the inner tank hinge are fixedly connected, the inner tank hinge and the ground are fixedly connected, and a rotational pair is set between the door hinge and the inner tank hinge.

[0170] The outer door subsystem simulation model is a simulation model mainly composed of the inner door, the outer door, and other components. MotionView and MotionSolve are used as the front and back processors respectively. Unit selection is carried out according to the force characteristics of each component. The inner door, the outer door, and other components are all simulated using rigid body elements, and the inner door, the outer door, and other components are fixedly connected, and the inner door is fixedly connected to the door hinge.

[0171] The cable spring subsystem simulation model is a simulation model mainly composed of a cable, a fixed pulley, a spring group, and other components. Unit selection is carried out according to the force characteristics of each component. The fixed pulley is simulated using a rigid body element, the cable is simulated using a nonlinear finite element element, and the spring is simulated using a spring element. One end of the cable is connected to the door hinge, the other end of the cable is connected to the spring, the spring is fixedly connected to the ground, and contact is defined between the cable and the fixed pulley.

[0172] The lock subsystem simulation model is a simulation model mainly composed of a door lock and an inner tank lock and other components. The inner tank lock is composed of a lock box, a tension spring, and 2 locking teeth. MotionView and MotionSolve are used as the front and back processors respectively. Unit selection is carried out according to the force characteristics of each component. Both the door lock and the inner tank lock are simulated using rigid body elements. The door lock is fixedly connected to the inner door, the inner tank lock is fixedly connected to the inner tank, the 2 locking teeth are respectively connected to both ends of the tension spring, each locking tooth rotates around a defined rotational pair, and contact is defined between the locking teeth, the lock box, and the door lock.

[0173] The seal subsystem simulation model can be composed of multiple Forces (e.g., 34). Using the macroscopic modeling method, the entire sealing strip is divided into 34 equal parts. The mechanical behavior of each sealing strip is described by a macroscopic constitutive model. According to past test experience, a two-segment linear line mathematical expression is used to describe the macroscopic constitutive model of the sealing strip. MotionView and MotionSolve are used as the pre- and post-processors respectively. Unit selection is carried out according to the force characteristics of each component. The acting force of the sealing strip is a pair of action and reaction forces. 34 Forces are used, and each Force acts on the inner container and the inner door respectively.

[0174] Furthermore, using the physical benchmarking data of each door core component, the non-destructive simulation models of each door core component are benchmarked one by one to obtain the normal behavior simulation model of each said door core component. Taking the non-destructive simulation model of any one of the door core components as an example, the following steps can be used to achieve this:

[0175] First, run the non-destructive simulation model to generate a virtual force time series curve. This virtual force time series curve describes the change in the predicted opening force of the non-destructive simulation model at different opening and closing angles.

[0176] Specifically, the Transient solver in MotionSolve can be used for solving. The outer door subsystem and the skeleton subsystem are connected by hinges. Set the rotating pair and the position of the outer door opening, and at the same time set the rotational motion of the door body. The analysis total time is 20s.

[0177] During the simulation process of running the non-destructive simulation model, the door body experiences the same time series actions. From the door body locked state, it opens to a preset angle and then returns to the original locked state. The simulation uses angle control, that is, the time used for opening 1° is the same for each opening and closing process of the non-destructive simulation model. Through the above simulation, the virtual force time series curve and the virtual angle time series curve of the virtual non-destructive simulation model can be obtained as the simulation output data. Since the angle control method is used and the angles are the same at the preset time series, only the virtual force time series curve needs to be concerned in the subsequent analysis.

[0178] Then, compare the virtual force time series curve generated by the non-destructive simulation model with the force time series curve sample obtained from the physical opening and closing experiment of this door core component, and calculate the curve difference between the two.

[0179] Optionally, the curve difference can be quantified by the mean square error (MSE) or other statistical indicators, which will not be elaborated here.

[0180] Furthermore, the model parameters of the non-destructive simulation model can be adjusted according to the calculated curve difference. Multiple operations can be iteratively carried out until the curve difference is within the preset range.

[0181] For example, if the mean square error between the virtual force time series curve and the force time series curve sample (i.e., the test curve) is less than a certain threshold (such as 0.01 N²), it is considered that the adjustment of the model parameters is completed. At this time, the virtual force time series curve output by the lossless simulation model is basically consistent with the force time series curve sample of the real physical prototype, that is, the model calibration is completed.

[0182] The adjusted lossless simulation model is the normal behavior simulation model, which can accurately simulate the behavior of the core components of the door body in a non-damaged state.

[0183] By using the above method to calibrate the lossless simulation models of each core component of the door body respectively, the normal behavior simulation models corresponding to each core component of the door body can be obtained.

[0184] Finally, by fusing the normal behavior simulation models of all the core components of the door body obtained by calibration, the complete door body system simulation model can be obtained in the present invention.

[0185] Taking the above dishwasher door body system as an example, by calibrating the outer door subsystem simulation model, the hinge subsystem simulation model, the pull rope spring subsystem simulation model, the lock button subsystem simulation model and the seal subsystem simulation model, the normal behavior simulation models of the corresponding complete door body system, hinge subsystem, pull rope spring subsystem, lock button subsystem and seal subsystem can be obtained. Finally, by fusing all these normal behavior simulation models in the simulation software, the door body system simulation model corresponding to the dishwasher door body system can be obtained.

[0186] The method for constructing the door body damage detection model provided by the present invention calibrates the lossless simulation models of each core component of the door body by using physical calibration data. By adjusting the model parameters to make the simulation output consistent with the experimental data, the normal behavior simulation model that can accurately reflect the real behavior of the door body is finally obtained. This process lays a solid foundation for the subsequent construction of the door body damage detection model and effectively improves the reliability of the model.

[0187] Based on the content of the above embodiments, as an alternative embodiment, the updating of the door body system simulation model according to the damage coefficient of each core component of the door body at each discrete time point in the preset time series to obtain the damage behavior simulation models in different damage states at each discrete time point includes:

[0188] Based on the current opening and closing times of any core component of the door body at any discrete time point and the fatigue opening and closing times limit value of the any core component of the door body, determine the damage coefficient of the any core component of the door body at the any discrete time point;

[0189] Determine a scaling coefficient of a target parameter related to the damage of any door core component based on the damage coefficient;

[0190] At any discrete time point, update the door system simulation model according to the scaling coefficients of the target parameters related to the damage of all door core components, and obtain the damage behavior simulation model at any discrete time point;

[0191] Traverse all the discrete time points to obtain all the damage behavior simulation models.

[0192] Through the description of the above embodiments, a normal behavior simulation model of the door system is established, which is idealized as intact throughout the time series. However, to construct a high-fidelity digital twin model, it is necessary to consider the damage evolution process of the real physical prototype. And how to consider the change of the structural health state of the door system is a technical difficulty in the industry.

[0193] The present invention proposes reasonable basic assumptions through the fatigue damage mechanics theory and establishes a theoretical relationship between the number of door openings and the structural damage evolution. The following continues to elaborate in detail on how to update the door system simulation model according to the damage coefficients of each door core component at each discrete time point in the preset time series to obtain the damage behavior simulation models in different damage states.

[0194] During the entire life cycle of the door system, the action time series it experiences is to open the door and then close the door, repeating this until its service life. In view of this, the present invention proposes several reasonable mechanical basic assumptions:

[0195] 1) Assume that for two consecutive door openings and closings, if the opening angles are the same, then the damage to each door core component caused by these two door openings and closings is the same. The present invention adopts a special experimental design to make the opening angle of each door opening and closing the same, so the damage of each door opening and closing is the same. Assume that the fatigue life of a certain door core component is N times, then the damage of each door opening and closing is 1 / N, which is represented by the D value.

[0196] 2) Assume that the damage of each door core component of the door system can be linearly accumulated. When the damage of a certain door core component accumulates to a certain extent (such as reaching a certain limit value), then this door core component reaches its service life. This limit value can be assumed to be 1 according to Miner's fatigue damage mechanics.

[0197] 3) For subsequent model order reduction sampling, curves with obvious differences in the relationship between the opening force and the opening angle need to be collected. Assume that after every m opening and closing operations, the relationship between the opening force and the opening angle changes significantly. Within m operations, assume that the change in the relationship between the opening force and the opening angle is not obvious and can be regarded as the same. Here, m can be a data such as 100 or 200.

[0198] Through the above assumptions, a connection can be established between the number of door opening operations and the process of structural damage evolution. The fatigue life of each core component of the door body under the experimental design can be obtained through simulation methods or experimental methods, thus perfectly solving the technical problem that damage evolution cannot be directly established by simulation software at the level of the dishwasher door body system.

[0199] During the construction process of the damage behavior simulation model, its object is the complete door body system simulation model. The Transient solver can be used for solving. The outer door subsystem and the skeleton are connected by hinges, and a rotational pair is set. The position of the outer door opening is set, and the rotational motion of the door body is set. The total analysis time is 20 s, etc. During the simulation process of the damage behavior simulation model in each damage state, the door body system experiences the same timing actions, from the door body locked state, opening to a preset angle, and then returning to the original locked state. The simulation also synchronously adopts angle control, that is, during each opening and closing simulation process, the time used to open the door by 1° is the same. Through the above simulation, the force time series curve of the virtual simulation prototype in each damage state can be obtained (since the angle control method is adopted, the angle time series curve is actually the same at this time, so it can be not considered), which is used as the output data of the damage behavior simulation model.

[0200] The steps to obtain the damage behavior simulation model in different damage states at each of the discrete time points can generally be divided into:

[0201] Step 1, for each core component of the door body, at each discrete time point, calculate the damage coefficient according to its current number of opening and closing operations and the fatigue opening and closing operation limit value. According to experimental data or theoretical analysis, determine the fatigue opening and closing operation limit value N lim of each core component of the door body, indicating that the core component of the door body may experience fatigue failure when reaching this number. Record the actual number of opening and closing operations N cuee of each core component of the door body during use, then the damage coefficient can be calculated according to the formula: D i = N cuee / N lim ; where D i represents the damage coefficient of the i-th core component of the door body, and its value range is [0, 1]. 0 indicates no damage, and 1 indicates complete failure.

[0202] Step 2: Determine the scaling factors of the target parameters related to door body damage according to the damage coefficients of each door body core component, including: determining the key parameters related to the damage of each door body core component, such as the stiffness of the pull rope spring, the stiffness of the latch spring, the resilience of the sealing strip, etc. According to the damage coefficients D of each door body core component at each discrete sampling point i , the scaling factors of the target parameters can be calculated.

[0203] For example, for the pull rope spring subsystem, the scaling factor of its pull rope stiffness can be expressed as: S i = 1 - D i , which means that as the damage coefficient increases, the target parameter (such as the pull rope stiffness) will decrease accordingly.

[0204] Step 3: At each discrete time point, update the door body system simulation model according to the scaling factors of the target parameters of all door body core components to obtain the damage behavior simulation model at this discrete time point. The specific steps include: for each door body core component, update the corresponding target parameter in the door body system simulation model according to the scaling factor S i of its target parameter. For example, for the pull rope spring subsystem, update its initial pull rope stiffness K0 to K updated : K updated = K0 S i . Finally, the updated target parameters can be used to run the door body system simulation model to obtain the damage behavior simulation model at this discrete time point and output the opening force time series data.

[0205] Step 4: Traverse all preset discrete time points, repeat the above steps, and obtain the damage behavior simulation models at each discrete time point. For example, starting from the first discrete time point, gradually traverse all discrete time points, and at each discrete time point, update the door body system simulation model according to the current damage coefficient and the scaling factor of the target parameter, and run the simulation. Collect the results of the damage behavior simulation models obtained at each discrete time point to form a complete set of damage behavior simulation models.

[0206] The present invention determines the damage coefficient based on the current opening and closing times and the fatigue opening and closing times limit value of each door body core component, and updates the target parameters in the door body system simulation model according to the damage coefficient, so as to obtain the damage behavior simulation models in different damage states. This process provides detailed damage state data for the subsequent construction of the door body damage detection model, ensuring that the model can accurately reflect the dynamic damage behavior of the door body system in different usage stages.

[0207] Based on the content of the above embodiments, as an alternative embodiment, the target parameters of the hinge subsystem include the deformation parameters at the connection point between the hinge and the pull rope. The target parameters of the pull rope spring subsystem include the pull rope stiffness and the telescopic spring stiffness. The target parameters of the latch subsystem include the latch spring stiffness. The target parameters of the sealing subsystem include the sealing strip stiffness.

[0208] Next, taking the dishwasher door system as an example, the damage conditions of its four subsystems will be described using corresponding damage coefficients. The damage coefficients of the entire door system simulation model include:

[0209] 1) Regarding the hinge damage coefficient D1 corresponding to the hinge subsystem:

[0210] According to the force analysis of the hinge, the damage is described by the deformation at the connection point between the hinge and the pull rope. Since the hinge is a rigid body in the model and does not deform, the damage is characterized by the change in the position of the connection point, that is, the deformation parameter at the connection point between the hinge and the pull rope.

[0211] Figure 4 It is a schematic diagram of the method for determining the hinge damage coefficient provided by the present invention. As Figure 4 shown, the vertical line represents the original position of the connection point between the hinge and the pull rope, and the inclined line represents the state position after the deformation of the connection point between the hinge and the pull rope. Then the deformation parameter of the connection point can be expressed as: R θ. Then the damage coefficient D1 = θ1 / θ0, 0 ≤ θ1 ≤ θ0, 0 ≤ D1 ≤ 1. R represents the radial distance from the hinge center to the pull rope connection point, which is used to convert the angular change θ into an actual displacement.

[0212] 2) Regarding the pull rope damage coefficient D2 corresponding to the pull rope spring subsystem:

[0213] Considering that the pull rope in the pull rope spring subsystem is composed of a steel wire rope wrapped with nylon, and its force is mainly borne by the steel wire rope, which is in a pure tension state. Therefore, in the present invention, the change in the pull rope stiffness is used to characterize a target parameter for judging damage in the pull rope spring subsystem.

[0214] Specifically, define the initial stiffness of the pull rope as K0, and its stiffness after multiple openings and closings at any discrete time point is K1. Then the damage coefficient D2 of this target parameter in the pull rope spring subsystem at this discrete time point can be defined as D2 = (K0 - K1) / K0. Where 0 ≤ K1 ≤ K0, 0 ≤ D2 ≤ 1.

[0215] 3) Regarding the pull rope damage coefficient D3 corresponding to the pull rope spring subsystem:

[0216] At the same time, the present invention also considers that the change in the stiffness of the telescopic spring in the rope spring subsystem is also an important target parameter reflecting the damage degree of the subsystem, and its corresponding damage formula can be expressed as: D3=(K2-K3) / K2, where 0≤K3≤K2, and 0≤D3≤1. The initial stiffness of the telescopic spring is K2, and its stiffness after multiple switches at any discrete time point is K3.

[0217] 4) Regarding the sealing strip damage coefficient D4 corresponding to the sealing subsystem:

[0218] The sealing strip only generates a certain amount of resilience when the door is opened, as it returns to its original shape. After the door is separated from the frame subsystem, the resilience of the sealing strip disappears. Therefore, the determination of the damage coefficient of the sealing subsystem is related to the specific damage behavior simulation model, and its most direct target parameter is the change in the stiffness of the sealing strip.

[0219] Figure 5 The following is a schematic diagram of the stiffness curve of the sealing strip provided by the present invention, wherein the horizontal axis represents the compression deformation of the sealing strip when subjected to an external force, usually expressed in length units (such as millimeters or inches), and the vertical axis represents the force applied to the sealing strip, usually expressed in force units (such as Newtons or pounds). Figure 5 As shown in the figure, the sealing strip stiffness curve describes the force response of the sealing strip under different compression deformations, reflecting the stiffness characteristics of the sealing strip. The slope of the stiffness curve (i.e. the ratio of force to compression deformation) reflects the stiffness of the sealing strip. The larger the slope of the stiffness curve, the harder the sealing strip; the smaller the slope of the stiffness curve, the softer the sealing strip.

[0220] Therefore, when performing the scaling coefficient of the target parameter related to the door body damage of the sealing subsystem at any discrete time point, the present invention quantifies the damage degree of the sealing strip and determines the corresponding damage coefficient D4, where D4=0 indicates that the sealing strip is not damaged, and D4=1 indicates that the sealing strip is completely ineffective.

[0221] Furthermore, the scaling factor of the sealing strip stiffness curve can be calculated according to the damage coefficient D4: S4=1-D4. For example, if D4=0.2, the scaling factor of the sealing strip stiffness curve=0.8.

[0222] Then, at each discrete time point, the sealing strip stiffness curve can be updated according to the scaling factor S4: K5 = K4 × S 4, where K4 is the initial stiffness curve slope of the sealing strip, and K5 is the updated stiffness curve slope.

[0223] By running the door system simulation model using the updated stiffness curve, a damage behavior simulation model at that discrete time point can be obtained.

[0224] 5) Regarding the spring damage coefficient D5 corresponding to the lock buckle subsystem:

[0225] At the initial moment of opening the door, the opening force is mainly the locking force generated by the lock buckle, which is mainly the pulling force caused by the deformation of the locking spring. Then, the damage of the entire lock buckle subsystem can be understood as mainly caused by the locking spring. Furthermore, the change in the stiffness of the locking spring in the door system simulation model can be defined to correlate with its corresponding damage coefficient.

[0226] Specifically, the initial stiffness of the locking spring can be defined as K6, and the stiffness after multiple openings and closings is K7. The damage coefficient D5 = (K6 - K7) / K6, where 0 ≤ K7 ≤ K6 and 0 ≤ D5 ≤ 1.

[0227] Based on the above description, the construction of a 3D parametric damage behavior simulation model considering damage evolution can be completed. A total of 5 target parameters D1 - D5 associated with each core component of the door body are set. At each discrete time point, according to the number of door openings, the corresponding damage coefficients D1 - D5 are calculated. Then, according to the above definitions of the damage coefficients, the door system simulation model is automatically updated to obtain corresponding damage behavior simulation models one by one.

[0228] The method for constructing the door body damage detection model provided by the present invention determines the damage coefficient based on the current number of openings and closings of each core component of the door body and the fatigue opening and closing times limit value, and updates the target parameters in the door system simulation model according to the damage coefficient, so as to obtain the damage behavior simulation models under different damage states. This process provides detailed damage state data for the subsequent construction of the door body damage detection model, ensuring that the model can accurately reflect the dynamic behavior of the door system at different usage stages.

[0229] Based on the content of the above embodiments, as an alternative embodiment, training the initial network model using any training data in the training dataset to obtain the door body damage detection model includes:

[0230] Sequentially inputting the training data corresponding to each discrete time point into the initial network model according to the preset time sequence to obtain an opening force prediction sequence output by the initial network model;

[0231] According to the network error between the opening force prediction sequence and the opening force label sequence, gradually update the model parameters of the initial network model until the model converges to obtain the door body damage detection model;

[0232] The opening force label sequence is composed of the opening forces at all discrete time points in the preset time sequence.

[0233] When detecting and calculating door body damage based on all the damage behavior simulation models obtained directly according to the methods mentioned in the above embodiments, there are limitations such as low computational efficiency, limited generalization ability, and complex data processing. In particular, these door body damage detection models are usually based on physical equations and complex mechanical calculations. Each time the simulation model runs, it requires a large amount of computational resources and time, and the amount of generated data is usually very large. Especially when considering multiple discrete time points and multiple damage states, processing and analyzing this data requires complex algorithms and a large amount of computational resources, and thus cannot be used as a digital twin model applicable to actual situations.

[0234] The present invention creatively uses a deep learning network structure algorithm to train the data generated by a 3D parametric damage behavior simulation model, and outputs a reduced-order damage behavior simulation model considering damage behavior as a digital twin model of the door body system considering damage evolution.

[0235] As an alternative embodiment, the present invention selects the Long Short-Term Memory (LSTM) as the basic architecture of the initial network model because LSTM can effectively process time series data and capture the time-dependent relationships in the input data. However, in actual application, other network models can also be considered as the basic model architecture for training to obtain a door body damage detection model. For example, a Recurrent Neural Network (RNN), a Gated Recurrent Unit (GRU), a Transformer network model, or in scenarios with higher detection accuracy requirements, a hybrid model composed of multiple network models can be used as the basic model architecture.

[0236] Figure 6 It is a schematic diagram of the training process of the door body damage detection model provided by the present invention. Next, it will be combined with Figure 6 As shown, taking the use of LSTM as the basic model architecture as an example, it will be detailedly described how to implement the training of the door body damage detection model. It mainly includes but is not limited to the following steps:

[0237] Step 1, based on the method provided in the foregoing embodiments, obtain the damage behavior simulation model corresponding to each discrete time point, so as to generate a training data set by using these damage behavior simulation models. Each data sample in the training data set includes input data and label data. Among them, the input data is the opening angle data at a discrete time point and the damage coefficients of each core component of the door body, and the corresponding label data is the corresponding opening force data.

[0238] Specifically, taking the door system of a general dishwasher as an example, its entire life cycle will run about 50,000 times. Every 1,000 times (depending on the actual selected sampling frequency), a simulation model of the door system is taken. According to the assumptions and methods provided in the above embodiments, the damage value coefficients of each subsystem (core components of the door) of the door system simulation model can be obtained each time. Based on the damage coefficients, a damage behavior simulation model corresponding to a door system simulation model can be obtained. In this way, 50 door system simulation models can be obtained. All the door system simulation models are input into MotionSolve for solution to obtain the opening force time series dataset of the dishwasher. Based on these opening force time series datasets, a training dataset for the improved LSTM deep learning neural network algorithm can be constructed.

[0239] Step 2: Select an LSTM network as the basic architecture of the initial network model. Its input layer can receive the opening angle and the damage coefficients of each core component of the door. The LSTM layer processes the time series data to extract time-dependent features, and its output layer outputs the predicted opening force.

[0240] Step 3: According to the preset time series, sequentially input the training data corresponding to each discrete time point into the initial network model, including the input data and label data recorded at each discrete time point from the training dataset, and input the input data (opening angle and damage coefficient) into the initial network model.

[0241] Run the initial network model to obtain the opening force prediction sequence output by the model, that is, use the initial network model to infer the input data and generate the opening force prediction sequence (collect the prediction results at each discrete time point to form a complete opening force prediction sequence).

[0242] Step 4: Calculate the network error according to the difference between the prediction sequence and the label sequence. Obtain the true opening force data at each discrete time point from the training dataset to form the opening force label sequence. Use the mean square error (MSE) or other statistical metrics to calculate the network error between the prediction sequence and the label sequence. For example:

[0243] ;

[0244] where N is the total number of discrete time points, and respectively represent the opening force error at the i-th discrete time point.

[0245] Step 5: According to the calculated network error, gradually update the parameters of the initial network model until the model converges. The specific steps include:

[0246] 1) Use the backpropagation algorithm to calculate the gradient of the error with respect to the model parameters.

[0247] 2) Use an optimization algorithm (such as Adam) to update the model parameters according to the gradient.

[0248] 3) Repeat the above process until the network error no longer decreases significantly and the model converges.

[0249] Step 6, through the above training process, the parameters of the initial network model are gradually optimized, and finally a high-precision door damage detection model is obtained. This door damage detection model can accurately predict the corresponding opening force according to the input opening angle and damage coefficient, so as to realize the real-time detection of the door damage state.

[0250] The method for constructing the door damage detection model provided by the present invention uses the training data set to train the initial network model, gradually optimizes the model parameters, and finally obtains a high-precision door damage detection model, which can significantly improve the calculation efficiency, enhance the generalization ability, simplify the data processing, and meet the requirements of real-time monitoring and adaptation to complex working conditions. It is a key step to realize efficient and accurate door damage detection.

[0251] Based on the content of the above embodiments, as an alternative embodiment, training the first initial network model with any training data in the training data set to obtain the door damage detection model includes:

[0252] According to the preset time sequence, sequentially input the training data corresponding to each discrete time point into the first initial network model to obtain an opening force prediction sequence output by the first initial network model;

[0253] According to the network error between the opening force prediction sequence and the opening force label sequence, gradually update the model parameters of the first initial network model until the model converges to obtain the door damage detection model;

[0254] The opening force label sequence is composed of the opening forces at all discrete time points in the preset time sequence.

[0255] Specifically, the training data set is generated based on the damage behavior simulation model, and the damage behavior simulation model considers the behaviors of the core components of the door (such as hinge subsystem, pull rope spring subsystem, lock button subsystem, and sealing subsystem, etc.) in different damage states. By running the damage behavior simulation model within the preset time sequence, the opening angle, damage coefficient, and corresponding opening force data at each discrete time point are generated.

[0256] Taking the opening angle at each discrete time point and the damage coefficients of each core component of the door as input data, and taking the opening force as the label data of this input data (defining the opening force label sequence as the sequence composed of the opening forces at all discrete time points in the preset time sequence), then the training data corresponding to each discrete time point can be obtained.

[0257] The first initial network model can adopt a long short-term memory network (LSTM) architecture, which can effectively process time series data and capture time dependencies, and is suitable for the complex mapping relationship between the opening angle, damage coefficient, and opening force in the door body damage detection task.

[0258] According to the preset time sequence, the training data corresponding to each discrete time point is sequentially input into the first initial network model. The input data includes the opening angle and the damage coefficients of each core component of the door body. The first initial network model outputs an opening force prediction sequence according to the input data. This opening force prediction sequence represents the predicted value of the opening force of the model at each discrete time point.

[0259] Compare the opening force prediction sequence output by the model with the true opening force label sequence, and calculate the network error. This network error can be measured by the mean square error (MSE). According to the calculated network error, update the parameters of the first initial network model through the backpropagation algorithm. Common optimization algorithms include Adam or RMSprop, etc., which can automatically adjust the learning rate and accelerate the convergence of the model.

[0260] Repeat the above training process until the network error no longer decreases significantly and the model converges. The judgment criterion for convergence can be that the error change in consecutive multiple training epochs is less than a preset threshold.

[0261] It should be noted that in the above training process, an early stopping mechanism can be adopted. When the error on the validation set no longer improves in consecutive multiple training epochs, stop the training to prevent overfitting. The generalization ability of the trained model can also be evaluated through cross-validation (such as k-fold cross-validation), and the optimal model parameters can be selected.

[0262] After the above training process, the parameters of the first initial network model are gradually optimized, and finally a high-precision door body damage detection model is obtained. This model can accurately predict the corresponding opening force according to the input opening angle and damage coefficient, so as to realize the real-time detection of the damage state of the door body.

[0263] The door body damage detection method provided by the present invention can efficiently train a high-precision door body damage detection model, providing a reliable technical means for the health management of the door body system.

[0264] Based on the content of the above embodiments, as an alternative embodiment, the obtaining of the door body damage recognition agent through the reinforcement learning of the door body damage detection model includes:

[0265] Construct a reinforcement training set using the door body damage detection model; any reinforcement training data in the reinforcement training set includes the opening angle and opening force at any discrete time point, and the label of any reinforcement training data is the damage coefficient of each door body core component at the any discrete time point, and the opening force is obtained by inputting the opening angle and the damage coefficient of each door body core component into the door body damage detection model;

[0266] Train the second initial network model using the reinforcement training set to obtain the door body damage recognition agent.

[0267] As an optional embodiment, the present invention provides a method for obtaining a door body damage recognition agent through reinforcement learning. This method uses a door body damage detection model to generate a reinforcement training set and trains a second initial network model based on this training set, and finally obtains a door body damage recognition agent that can identify the damage state of the door body in real time. The method mainly includes but is not limited to the following steps:

[0268] First of all, it is necessary to construct a reinforcement training set.

[0269] The door body damage detection model is constructed based on deep learning and digital twin technology and can simulate the behavior of the door body system in different damage states. This door body damage detection model outputs the corresponding opening force by inputting the opening angle and the damage coefficient of each door body core component, and this model provides a basis for constructing the reinforcement training set.

[0270] Optionally, the steps for generating a reinforcement training set using the door body damage detection model specifically include:

[0271] 1) Select multiple discrete time points within a preset time sequence, and these time points cover the whole process of the door body system from no damage to approaching the service life.

[0272] 2) For each discrete time point, generate the corresponding opening angle and the damage coefficient of each door body core component as the input part of the reinforcement training data.

[0273] 3) Input the opening angle and the damage coefficient into the door body damage detection model, and the model outputs the corresponding opening force as the label part of the reinforcement training data.

[0274] In this way, a reinforcement training data can be obtained for each discrete time point. Its input is the opening angle and the damage coefficient of each door body core component, and the corresponding label is the opening force at the same discrete time point. The reinforcement training set consists of multiple reinforcement training data, and each reinforcement training data corresponds to a discrete time point.

[0275] Train the second initial network model using the obtained enhanced training dataset so that it can learn the complex mapping relationship among the opening angle, damage coefficient, and opening force, specifically including:

[0276] According to the preset time sequence, sequentially input each enhanced training data in the enhanced training dataset into the second initial network model. The input data includes the opening angle and the damage coefficients of each core component of the door body. The second initial network model outputs a predicted opening force sequence based on the input data.

[0277] Compare the predicted opening force output by the second initial network model with the label in the enhanced training data, calculate the network error, and update the model parameters of the second initial network model through the backpropagation algorithm according to the calculated network error. Repeat the above process until the network error no longer significantly decreases and the model converges, and the second initial network model is optimized into a door body damage recognition agent.

[0278] As an alternative embodiment, both the above first initial network model and the second initial network model are obtained based on the long short-term memory network model as the basic model architecture.

[0279] The door body damage detection method provided by the present invention can efficiently train a door body damage recognition agent, providing a reliable technical means for the health management of the door body system.

[0280] Figure 7 It is a schematic structural diagram of the door body damage detection device provided by the present invention, as Figure 7 shown, mainly including but not limited to:

[0281] An agent training unit 71, configured to obtain a door body damage recognition agent through reinforcement learning of the door body damage detection model;

[0282] An agent recognition unit 72, configured to use the door body damage recognition agent to recognize the opening angle sequence and the opening force sequence of the door body to be measured within a preset time sequence, and output the damage coefficient sequence of each core component of the door body to be measured within the preset time sequence;

[0283] A door body detection unit 73, configured to determine the detection result of the door body to be measured based on the damage coefficient sequence of each core component of the door body.

[0284] It should be noted that when the door body damage detection device provided by the present invention is specifically operating, it can implement the door body damage detection method provided in any of the above embodiments, which will not be elaborated here one by one.

[0285] The door body damage detection device provided by the present invention, after establishing a 3D parametric damage behavior simulation model considering damage, reduces the order of the discrete damage behavior simulation model through deep learning to obtain a door body damage detection model, and then learns the door body damage detection model considering damage evolution behavior based on a reinforcement learning algorithm to obtain a door body damage recognition agent, which can realize real-time monitoring of the health state of the door body system.

[0286] In order to fully illustrate the advantages of the construction method of the door body damage detection model provided by the present invention in actual detection, the following will be described in detail in combination with relevant test data.

[0287] Figure 8 It is one of the schematic diagrams of the performance of the door body damage detection model provided by the present invention in the training dataset. Figure 9 It is the second schematic diagram of the performance of the door body damage detection model provided by the present invention in the training dataset. Figure 10 It is the third schematic diagram of the performance of the door body damage detection model provided by the present invention in the training dataset. Figure 11 It is the fourth schematic diagram of the performance of the door body damage detection model provided by the present invention in the training dataset. Figure 12 It is the fifth schematic diagram of the performance of the door body damage detection model provided by the present invention in the training dataset. Among them, the abscissa represents the time step, each unit corresponds to 0.025 seconds. For example, 400 on the abscissa represents 10 seconds; the ordinate represents the numerical values of the predicted value and the actual value, such as the opening force, and the unit is Newton (N). Among them, one of the two force time series curves represents the predicted value, and the other represents the actual value (the dotted line represents the predicted value, and the solid line represents the actual value).

[0288] Among them, during Figures 8 - 12 the training process of the initial network model shown, the total number of door openings (the data scale in the training dataset) is 20900, 27100, 34800, 4200, and 9700 respectively. Mainly through the comparison relationship between the predicted value and the actual value, the abscissa also represents the time step, the ordinate represents the opening force, and the force time series curve fully demonstrates the convergence of the door body damage detection model during the training process, which can effectively prove that the door body damage detection model obtained through a finite number of trainings has a high prediction accuracy.

[0289] Figure 13 It is one of the schematic diagrams of the performance of the door body damage detection model provided by the present invention in the test dataset. Figure 14 It is the second schematic diagram of the performance of the door body damage detection model provided by the present invention in the test dataset. Figure 15 It is the third schematic diagram of the performance of the door body damage detection model provided by the present invention in the test dataset. Figure 16 It is the fourth schematic diagram of the performance of the door body damage detection model provided by the present invention in the test dataset. Figure 17It is the fifth schematic diagram of the performance of the door body damage detection model provided by the present invention on the test data set. Among them, the abscissa of the coordinate system where the force time series curve is located represents the time step, and the ordinate represents the value of the door opening force.

[0290] Among them, during the training process of the initial network model shown in Figures 13 - 17 the total number of door openings (the data scale in the test data set) is 10400, 11400, 12200, 12900, and 13600 respectively. The convergence of the door body damage detection model during the test is mainly shown through the comparison relationship between the predicted value and the actual value in the two force time series curves, which can help evaluate the generalization ability and prediction accuracy of the model on unseen data. Through the verification of the test data set, it can be determined that the door body damage detection model obtained by the present invention has good generalization and high prediction accuracy.

[0291] Figure 18 It is one of the schematic diagrams of the evolution of the loss function value of the door body damage detection model provided by the present invention during the training process. Figure 19 It is the second schematic diagram of the evolution of the loss function value of the door body damage detection model provided by the present invention during the training process. Among them, the abscissa represents the number of iterations, and the ordinate represents the loss function value. The curve is mainly used to represent the change of the loss function value as the number of training iterations increases. If the loss function value drops rapidly and tends to be stable, it indicates that the model converges quickly. The fact that the loss function value finally tends to a small value indicates that the training effect of the model is good.

[0292] Refer to Figures 18 - 19 As shown, the Loss error of the loss function value of the door body damage detection model reaches 2.53 at the 171st step, and the convergence accuracy is very high. Thus, a deep learning network model with very high fitting accuracy is obtained, which is used as the digital twin model of the damage behavior of the dishwasher door body system.

[0293] The above experiments fully prove that the construction method of the door body damage detection model provided by the present invention obtains an LSTM deep learning network model with very high fitting accuracy, and has the ability to effectively utilize temporal features to solve the problem of long-term temporal prediction.

[0294] Figure 20 It is the structural schematic diagram of the electronic device provided by the present invention, as shown in Figure 20As shown in the figure, the electronic device may include: a processor 2010, a communications interface 2020, a memory 2030, and a communication bus 2040. Among them, the processor 2010, the communications interface 2020, and the memory 2030 complete their mutual communication through the communication bus 2040. The processor 2010 may call the logical instructions in the memory 2030 to execute the door body damage detection method, which includes: obtaining a door body damage recognition agent through the reinforcement learning of the door body damage detection model; using the door body damage recognition agent to recognize the opening angle sequence and opening force sequence of the door body to be tested within a preset time sequence, and outputting the damage coefficient sequence of each core component of the door body to be tested within the preset time sequence; and determining the detection result of the door body to be tested based on the damage coefficient sequence of each core component of the door body.

[0295] In addition, when the logical instructions in the above-mentioned memory 2030 can be implemented in the form of software functional units and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk, or an optical disc that can store program codes.

[0296] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the door body damage detection method provided in the above-mentioned various embodiments. The method includes: obtaining a door body damage recognition agent through the reinforcement learning of the door body damage detection model; using the door body damage recognition agent to recognize the opening angle sequence and opening force sequence of the door body to be tested within a preset time sequence, and outputting the damage coefficient sequence of each core component of the door body to be tested within the preset time sequence; and determining the detection result of the door body to be tested based on the damage coefficient sequence of each core component of the door body.

[0297] In another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is configured to execute the door body damage detection method provided in the above-mentioned embodiments. The method includes: through reinforcement learning of the door body damage detection model, obtaining a door body damage recognition agent; using the door body damage recognition agent to identify the opening angle sequence and opening force sequence of the door body to be measured within a preset time sequence, and outputting a damage coefficient sequence of each core component of the door body to be measured within the preset time sequence; based on the damage coefficient sequence of each core component of the door body, determining the detection result of the door body to be measured.

[0298] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative effort.

[0299] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0300] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting door body damage, characterized in that, Including: Obtaining a door body damage recognition agent through reinforcement learning of a door body damage detection model; Using the door body damage recognition agent to identify the opening angle sequence and opening force sequence of the door body to be tested within a preset time sequence, and outputting the damage coefficient sequence of each door body core component of the door body to be tested within the preset time sequence; Determining the detection result of the door body to be tested based on the damage coefficient sequence of each door body core component; The door body damage detection model is constructed by the following steps: Based on the physical calibration data and the damage coefficients of each door body core component at each discrete time point in the preset time sequence, constructing a damage behavior simulation model for each discrete time point; the physical calibration data is obtained by performing physical opening and closing experiments on each door body core component of a real physical door body; Using a training data set to train a first initial network model to obtain the door body damage detection model; any training data in the training data set includes the opening angle of a damage behavior simulation model at its corresponding discrete time point and the damage coefficients of each door body core component, and the label of the training data is the opening force at the discrete time point.

2. The door body damage detection method according to claim 1, characterized in that The constructing a damage behavior simulation model for each discrete time point based on the physical calibration data and the damage coefficients of each door body core component at each discrete time point in the preset time sequence includes: Using the physical calibration data to perform calibration on each non-damaged simulation model one by one to obtain a normal behavior simulation model for each door body core component; Combining the normal behavior simulation models to form a door body system simulation model; According to the damage coefficients of each door body core component at each discrete time point in the preset time sequence, updating the door body system simulation model to obtain the damage behavior simulation models in different damage states at each discrete time point.

3. The door body damage detection method according to claim 2, wherein, The physical calibration data includes force time sequence curve samples obtained by performing physical opening and closing experiments on each door body core component under angle control respectively; The using the physical calibration data to perform calibration on each non-damaged simulation model one by one to obtain a normal behavior simulation model for each door body core component includes: For the non-damaged simulation model of any door body core component, adjusting the model parameters of the non-damaged simulation model according to the curve difference between the virtual force time sequence curve output by the non-damaged simulation model and the force time sequence curve sample until the curve difference is within a preset range, and obtaining the adjusted non-damaged simulation model as the normal behavior simulation model.

4. The door body damage detection method according to any one of claims 2-3, characterized in that, The door body core components include a complete door body system and all key subsystems, and the key subsystems include one or more of a hinge subsystem, a cable spring subsystem, a lock subsystem, and a sealing subsystem; the hinge subsystem is composed of a skeleton subsystem and an outer door subsystem; Correspondingly, the non-damaged simulation models include a complete door body system simulation model, a hinge subsystem simulation model, a cable spring subsystem simulation model, a lock subsystem simulation model, and a sealing subsystem simulation model.

5. The door body damage detection method according to claim 4, wherein, The combining the normal behavior simulation models to form a door body system simulation model includes: Integrate the normal behavior simulation models of the complete door system, hinge subsystem, cable spring subsystem, lock subsystem, and sealing subsystem corresponding to the complete door system simulation model, hinge subsystem simulation model, cable spring subsystem simulation model, lock subsystem simulation model, and sealing subsystem simulation model to obtain the door system simulation model.

6. The door body damage detection method according to claim 5, wherein Update the door system simulation model according to the damage coefficients of each door core component at each discrete time point in the preset time sequence to obtain the damage behavior simulation models in different damage states at each discrete time point, including: Based on the current opening and closing times of any door core component at any discrete time point and the fatigue opening and closing times limit value of the any door core component, determine the damage coefficient of the any door core component at the any discrete time point; Based on the damage coefficient, determine the scaling coefficient of the target parameter related to the door damage of the any door core component; At the any discrete time point, update the door system simulation model according to the scaling coefficients of the target parameters related to the door damage of all door core components to obtain the damage behavior simulation model at the any discrete time point; Traverse all the discrete time points to obtain all the damage behavior simulation models.

7. The door body damage detection method according to claim 6, characterized in that, The target parameters of the hinge subsystem include the deformation parameters at the connection point of the hinge and the cable, the target parameters of the cable spring subsystem include the cable stiffness and the telescopic spring stiffness, the target parameters of the lock subsystem include the locking spring stiffness, and the target parameters of the sealing subsystem include the resilience of the sealing strip.

8. The door body damage detection method according to claim 1, wherein Train the first initial network model using any training data in the training dataset to obtain the door damage detection model, including: According to the preset time sequence, sequentially input the training data corresponding to each discrete time point into the first initial network model to obtain the opening force prediction sequence output by the first initial network model; According to the network error between the opening force prediction sequence and the opening force label sequence, gradually update the model parameters of the first initial network model until the model converges to obtain the door damage detection model; The opening force label sequence is composed of the opening forces at all discrete time points in the preset time sequence.

9. The door body damage detection method according to claim 1, characterized in that, Obtain the door damage recognition agent through the reinforcement learning of the door damage detection model, including: Construct a reinforcement training set using the door damage detection model; any reinforcement training data in the reinforcement training set includes the opening angle and opening force at any discrete time point, and the label of the any reinforcement training data is the damage coefficient of each door core component at the any discrete time point, and the opening force is obtained after inputting the opening angle and the damage coefficients of each door core component into the door damage detection model; Train the second initial network model using the reinforcement training set to obtain the door damage recognition agent.

10. The door body damage detection method according to claim 9, characterized in that, The training the second initial network model using the reinforcement training set to obtain the door damage recognition agent includes: According to the preset time sequence, the reinforcement training data corresponding to each of the discrete time points is sequentially input into the second initial network model, and a damage coefficient prediction sequence output by the second initial network model is obtained; Based on the network error between the damage coefficient prediction sequence and the damage coefficient label sequence, the model parameters of the second initial network model are gradually updated until the model converges, and the door body damage recognition agent is obtained; The damage coefficient label sequence is composed of the damage coefficients of each core component of the door body at all discrete time points in the preset time sequence.

11. The door body damage detection method according to claim 10, wherein Both the first initial network model and the second initial network model are obtained based on the long short-term memory network model as the basic model architecture.

12. A door body damage detection device, characterized in that It includes: An agent training unit, configured to obtain a door body damage recognition agent through reinforcement learning of a door body damage detection model; An agent recognition unit, configured to use the door body damage recognition agent to recognize the opening angle sequence and the opening force sequence of a to-be-detected door body within a preset time sequence, and output a damage coefficient sequence of each core component of the to-be-detected door body within the preset time sequence; A door body detection unit, configured to determine the detection result of the to-be-detected door body based on the damage coefficient sequence of each core component of the door body; the door body damage detection model is constructed by the following steps: based on the physical calibration data and the damage coefficients of each core component of the door body at each discrete time point in the preset time sequence, a damage behavior simulation model of each discrete time point is constructed; the physical calibration data is obtained by performing a physical opening and closing experiment on each core component of a real physical door body; the first initial network model is trained using a training data set to obtain the door body damage detection model; any training data in the training data set includes an opening angle of a damage behavior simulation model at its corresponding discrete time point and the damage coefficients of each core component of the door body, and the label of the training data is the opening force at the discrete time point.

13. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the door body damage detection method according to any one of claims 1 to 11.

14. A non-transitory computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by the processor, it implements the door body damage detection method according to any one of claims 1 to 11.

15. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the door body damage detection method according to any one of claims 1 to 11.

Images