BIM nervous system architecture for rail transit

Abstract

The invention discloses a BIM (Building Information Modeling) nervous system architecture for rail transit, the BIM nervous system architecture comprises a data and model combination module, a control execution module and an animation linkage module, the data and model combination module comprises a station equipment three-dimensional model, a station basic three-dimensional model and a neural network model, the station equipment three-dimensional model is put into the station basic three-dimensional model through a position matching algorithm, the position matching algorithm is realized through the neural network model, the neural network model is a multi-layer perceptron, and the multi-layer perceptron is composed of an input layer, two hidden layers and an output layer. The rail transit comprehensive monitoring system has the advantages that the intelligent level of the rail transit comprehensive monitoring system is remarkably improved, the intuition of equipment monitoring and the convenience of operation are improved, and more efficient technical support is provided for operation management of rail transit.

Description

Technical Field

[0001] The present invention relates to the technical field of integrated monitoring of rail transit, and in particular to a BIM nervous system architecture for rail transit. Background Art

[0002] With the development of rail transit and the promotion of smart city construction, the management and operation and maintenance of stations increasingly rely on advanced technical means. The application of three-dimensional modeling technology (BIM, Building Information Modeling) in the field of rail transit provides an intuitive display of equipment layout and real-time status monitoring for the design, construction and management of stations. However, there are still many limitations in the actual application of existing station three-dimensional systems.

[0003] The traditional integrated monitoring system is limited to only displaying the operating status of equipment in the initial model on the three-dimensional model, unable to directly control the equipment in the model, nor directly expand and modify the model. This leads to the complexity and inconvenience of operation. When the station is renovated and the equipment configuration changes, the update and maintenance process of the existing three-dimensional model is very cumbersome. If new, deleted or modified equipment is required, such as adding a turnstile or a camera, usually the entire model needs to be reprocessed and then data binding is carried out, which is extremely time-consuming and laborious. There is another way. When the equipment configuration changes, only update the equipment list on the platform and do not reflect it in the three-dimensional model. Although this reduces the workload of model update, the newly added, deleted or modified equipment cannot be visually displayed in the model, which may lead to a big difference between the three-dimensional model of the station after 10 or 20 years and the actual station. The scalability and generalization of the traditional three-dimensional model are too poor, and it is easy to cause the model built with time and effort to be finally abandoned. Summary of the Invention

[0004] The purpose of the present invention is to provide a BIM nervous system architecture for rail transit according to the deficiencies of the above-mentioned prior art. The BIM nervous system architecture is composed of a data and model combination module, a control execution module and an animation linkage module. The data and model combination module realizes the deep integration of data and the three-dimensional model. The control execution module allows operators to directly perform real-time control on on-site equipment through the three-dimensional model. The animation linkage module ensures that when the equipment performs an action, the equipment in the three-dimensional model can display the corresponding animation effect, realizing the real-time synchronization of the equipment action and the model state, solving the problems that in the application of the traditional integrated monitoring in the three-dimensional model, the equipment cannot be directly controlled, the model cannot be expanded or directly modified, etc., significantly improving the intelligent level of the rail transit integrated monitoring system, enhancing the intuitiveness of equipment monitoring and the convenience of operation, and providing more efficient technical support for the operation and management of rail transit.

[0005] The object of the present invention is achieved by the following technical solutions:

[0006] A BIM nervous system architecture for rail transit, the BIM nervous system architecture includes a data and model combination module, the data and model combination module includes a three-dimensional model of station equipment, a three-dimensional model of station foundation and a neural network model, the three-dimensional model of station equipment is placed into the three-dimensional model of station foundation through a position matching algorithm, the position matching algorithm is implemented through the neural network model, the neural network model is a multi-layer perceptron, and the multi-layer perceptron is composed of an input layer, two hidden layers and an output layer;

[0007] The position matching algorithm includes the following steps:

[0008] S1: Obtain the feature X of the three-dimensional model of station equipment device and the feature X of the three-dimensional model of station foundation base and perform preprocessing;

[0009] S2: Use the neural network model to calculate the position parameters of the equipment in the three-dimensional model of station foundation

[0010] S2.1: Input the feature X of the three-dimensional model of station equipment device and the feature X of the three-dimensional model of station foundation base into the input layer and calculate the input feature vector X;

[0011] S2.2: Process the input feature vector X through the two hidden layers to obtain the output H1 of the first hidden layer and the output H2 of the second hidden layer in sequence;

[0012] S2.3: Use the output layer to calculate the position parameters of the equipment in the three-dimensional model of station foundation

[0013] S3: Train the neural network model using the loss function L;

[0014] S4: Calculate the gradients of the loss function L with respect to the weights and biases of the neural network model;

[0015] S5: Use the Adam optimization algorithm to update the weights and biases of the neural network model;

[0016] S6: Repeat steps S2 to S5. Through multiple iterations, make the loss function L reach a convergence state, and the weights and biases of the neural network model are adjusted to the position that minimizes the loss function, so that the neural network model can accurately adjust the position parameters of the equipment;

[0017] S7: Using the trained neural network model, given the feature X of the 3D model of the station equipment device and the feature X of the 3D model of the station foundation base , the neural network model outputs the position parameters of the equipment in the 3D model of the station foundation

[0018] In step S1 of the position matching algorithm,

[0019] the feature X of the 3D model of the station equipment device includes the geometric shape, size, and predefined spatial information of the equipment, and the feature X of the 3D model of the station foundation base includes the structure and spatial information of the building.

[0020] In step S2.1 of the position matching algorithm,

[0021] The calculation formula for the input feature vector X is:

[0022] X = [X device , X base ;

[0023] where X is a vector of length d device + d base , d device is the dimension of the feature of the 3D model of the station equipment, and d base is the dimension of the feature of the 3D model of the station foundation.

[0024] In step S2.2 of the position matching algorithm,

[0025] The calculation formula for the output H1 of the first hidden layer is:

[0026] H1 = f(W1X + b1);

[0027] where W1 is the weight matrix of the first hidden layer, with dimension h1 × (d device + d base ); b1 is the bias vector of the first hidden layer, with dimension h1; f is the ReLU activation function;

[0028] The calculation formula for the output H2 of the second hidden layer is:

[0029] H2 = f(W2X + b2);

[0030] where W2 is the weight matrix of the second hidden layer, with dimension h1 × h2; b2 is the bias vector of the second hidden layer, with dimension h2.

[0031] In step S2.3 of the position matching algorithm,

[0032] The position parameters of the device in the basic three-dimensional model of the station The calculation formula is:

[0033]

[0034] In the formula, W3 is the weight matrix of the output layer, with dimension d ouput ×h2; b3 is the bias vector of the output layer, with dimension d output ; The dimension is d output , if d output = 12, where 9 parameters represent the 9 elements of the rotation matrix R, the rotation matrix R is a 3×3 matrix, and 3 parameters represent the 3 elements of the translation vector T, the translation vector T is a 3×1 vector, including three components (x, y, z), representing the translation distances along the x-axis, y-axis, and z-axis respectively.

[0035] In step S3 of the position matching algorithm

[0036] The loss function L measures the difference between the predicted value and the true value Y, and the loss function L uses the mean square error:

[0037]

[0038] In the formula, n is the number of training samples, Y i is the true position parameter of the i-th sample, is the predicted position parameter of the i-th sample.

[0039] In step S4 of the position matching algorithm

[0040] The gradient of the loss function L with respect to the weight matrix W The calculation formula is:

[0041]

[0042] The gradient of the loss function L with respect to the bias vector b The calculation formula is:

[0043]

[0044] In step S5 of the position matching algorithm

[0045] Initialize the first-order momentum m and the second-order momentum v:

[0046]

[0047] where β1 and β2 are attenuation coefficients used to control the attenuation rate of historical information, usually set to 0.9 and 0.999; θ t is the weight / bias of the current model;

[0048] Perform bias correction:

[0049]

[0050] Update the weight / bias of the neural network model:

[0051]

[0052] where ∈ is a small constant to prevent division by zero error, usually set to 10 -8 ; α is the learning rate that controls the step size of updating the weight / bias of the neural network model; t is the time step that tracks the number of current iterations, and it increases by 1 for each iteration; and are both bias correction terms.

[0053] The BIM nervous system architecture further includes a control and execution module, which includes an intelligent control module and a real-time feedback module. The intelligent control module is used to control the operating state of the device, and the real-time feedback module is used to feedback the operating state of the device.

[0054] The BIM nervous system architecture further includes an animation linkage module, which includes a key-frame animation module and a skeletal animation module. The key-frame animation module is used to simulate the basic actions of the device, and the skeletal animation module is used to simulate the movement of complex devices.

[0055] The advantages of the present invention are:

[0056] (1) The integrated monitoring system can achieve multi-dimensional control and status display. It not only retains the functions of the integrated monitoring system for control and status monitoring on the two-dimensional interface, but also can perform control on the visual three-dimensional interface. The station equipment is represented by a series of English letters and numbers in the system. The operation personnel no longer need to rely solely on experience and memory to operate the station. They can identify the equipment to be controlled according to the location of the equipment and the model icon, which is intuitive and efficient;

[0057] (2) The equipment model and the basic model of the station are decoupled. When the equipment configuration changes, modification of the equipment components is allowed. The flexibility and scalability of the model are high, and it can be customized and adjusted according to the change of requirements without affecting the unmodified model; The three-dimensional model can be independently edited, so the generalization ability of the model is strong. In the future, if new functions or technologies need to be integrated, there is no need to make large-scale modifications to the entire model, reducing the cost of model maintenance and improving the reusability;

[0058] (3) The animation linkage simulates the actions of real devices. When an operator executes a control command, the operator can see the response actions of the devices on the model in real time. This real-time feedback not only enhances the operator's interactive experience but also helps the operator more quickly understand the effects of control commands and the responses of the devices; it can monitor the operating status and actions of the devices more precisely, reduce the risk of misoperation, and improve operation safety.

[0059] (4) The integrated monitoring system is not only applicable to daily operations but also can be used for technical training and demonstrations; the 3D model can more comprehensively display the position and dynamic changes of the devices in space. Through 3D vision and animation linkage, new employees can more quickly understand, learn, and master device operation skills, reduce training costs, and at the same time, it can also be used to show technical capabilities and device functions to the outside world. Description of the Drawings

[0060] Figure 1 It is the overall schematic diagram of the BIM nervous system architecture of the present invention;

[0061] Figure 2 It is the working schematic diagram of the neural network model of the present invention;

[0062] Figure 3 It is the working schematic diagram of the control execution module of the present invention. Detailed Embodiments

[0063] The features of the present invention and other related features are further described in detail below with reference to the drawings through embodiments for the understanding of those skilled in the same industry:

[0064] Embodiment: As Figure 1 shown, this embodiment relates to a BIM nervous system architecture for rail transit. The BIM nervous system architecture mainly includes a data and model combination module, a control execution module, and an animation linkage module. The data and model combination module includes a 3D model of station equipment, a 3D model of station foundation, and a neural network model. The 3D model of station equipment is placed into the 3D model of station foundation through a position matching algorithm (combined with station drawings), and the position matching algorithm is implemented through the neural network model. The neural network model is a multi-layer perceptron, and the multi-layer perceptron consists of an input layer, two hidden layers, and an output layer. Specifically, the user can drag and drop the device icons (3D model of station equipment) to the 3D model of station foundation in the 3D editing configuration interface. When the station equipment changes, the device icons can be directly updated in the 3D editing configuration interface; after the user completes the drag-and-drop operation of the device icons and confirms, the system will save the newly added device information to the database or configuration file and automatically place the device icons at the corresponding positions on-site. When the user opens the model next time, the added devices and their positions can be directly seen, thus completing the save and synchronization.

[0065] The control execution module includes an intelligent control module and a real-time feedback module. The intelligent control module is used to control the operating state of the device, and the real-time feedback module is used to feedback the operating state of the device. Specifically, when a control instruction is issued by clicking on the device icon, spatial information is attached, so the command is sent to the device at the specific location corresponding to the station. This matching process is automatically executed, and the operating state of the device can be reflected by the icon of the device's 3D model. A corresponding icon is preset for each state of the device, and the icon corresponding to the current operating state of the device is displayed.

[0066] The animation linkage module includes a key-frame animation module and a skeletal animation module. The key-frame animation module is used to simulate the basic movements of the device, such as the translation of the train, the lifting of the vertical elevator, the rotation of the fan, etc. The skeletal animation module is used to simulate the movement of complex devices, such as the up and down movement of the escalator.

[0067] In this embodiment, the position matching algorithm includes the following steps:

[0068] S1: Obtain the feature X of the 3D model of the station equipment device and the feature X of the basic 3D model of the station base and perform preprocessing.

[0069] Among them, the feature X of the 3D model of the station equipment device includes the geometric shape, size and predefined spatial information of the equipment, and the feature X of the basic 3D model of the station base includes the structure and spatial information of the building.

[0070] S2: Use the neural network model to calculate the position parameters of the equipment in the basic 3D model of the station

[0071] S2.1: Input the feature X of the 3D model of the station equipment device and the feature X of the basic 3D model of the station base into the input layer and calculate the input feature vector X.

[0072] Among them, the calculation formula of the input feature vector X is:

[0073] X = [X device , Xx ase ;

[0074] In the formula, X is a vector with a length of d device +d base , d device is the dimension of the feature of the 3D model of the station equipment, and d base is the dimension of the feature of the basic 3D model of the station.

[0075] S2.2: Process the input feature vector X through two hidden layers to obtain the output H1 of the first hidden layer and the output H2 of the second hidden layer in sequence.

[0076] Among them, the hidden layer uses a linear transformation, including a weight matrix, a bias vector, and an activation function to calculate the output.

[0077] The calculation formula for the output H1 of the first hidden layer is:

[0078] H1 = f(W1X + b1);

[0079] In the formula, W1 is the weight matrix of the first hidden layer, with a dimension of h1×(d device +d base ); b1 is the bias vector of the first hidden layer, with a dimension of h1; f is the ReLU activation function;

[0080] The calculation formula for the output H2 of the second hidden layer is:

[0081] H2 = f(W2X + b2);

[0082] In the formula, W2 is the weight matrix of the second hidden layer, with a dimension of h1×h2; b2 is the bias vector of the second hidden layer, with a dimension of h2.

[0083] S2.3: Use the output layer to calculate the position parameters of the device in the basic 3D model of the station

[0084] Among them, the position parameters of the device in the basic 3D model of the station The calculation formula is:

[0085]

[0086] In the formula, W3 is the weight matrix of the output layer, with a dimension of d output ×h2; b3 is the bias vector of the output layer, with a dimension of d output ; The dimension of is d output , if d output = 12, 9 of which represent the 9 elements of the rotation matrix R. The rotation matrix R (describing how the device rotates in 3D space) is a 3×3 matrix, and 3 represent the 3 elements of the translation vector T. The translation vector T (describing how the device translates in 3D space) is a 3×1 vector, including three components (x, y, z), representing the translation distances along the x-axis, y-axis, and z-axis respectively.

[0087] S3: Use the loss function L to train the neural network model.

[0088] Among them, the loss function L measures the predicted value For the difference from the true value Y, the loss function L uses the Mean Squared Error (MSE):

[0089]

[0090] where n is the number of training samples, Y is the true position parameter of the i-th sample, and is the predicted position parameter of the i-th sample (the predicted position where the device is placed in the 3D model of the station foundation).

[0091] S4: Calculate the gradients of the loss function L with respect to the weights and biases of the neural network model.

[0092] Among them, the gradient of the loss function L with respect to the weight matrix W is calculated as follows:

[0093]

[0094] The gradient of the loss function L with respect to the bias vector b is calculated as follows:

[0095]

[0096] S5: Use the Adam optimization algorithm to update the weights and biases of the neural network model.

[0097] Among them, the core idea of gradient descent is to update the parameters along the negative gradient direction of the loss function to gradually reduce the loss value. Adam (Adaptive Moment Estimation) is a variant of gradient descent. Using the Adam optimization algorithm to update the weights and biases, it combines the advantages of momentum and RMSProp algorithms and has good performance.

[0098] Initialize the first-order momentum m and the second-order momentum v:

[0099]

[0100] where β1 and β2 are decay coefficients used to control the decay rate of historical information, usually set to 0.9 and 0.999; θ t is the weight / bias of the current model;

[0101] Perform bias correction:

[0102]

[0103] Update the weights / biases of the neural network model:

[0104]

[0105] where ∈ is a small constant to prevent division by zero error, usually set to 10 -8 ; α is the learning rate, which controls the step size of updating the weights / biases of the neural network model; t is the time step, which tracks the number of current iterations, and increases by 1 for each iteration; and are both bias correction terms.

[0106] S6: Repeat steps S2 to S5. Through multiple iterations, make the loss function L reach the convergence state. The weights and biases of the neural network model are adjusted to the position that minimizes the loss function, enabling the neural network model to accurately adjust the position parameters of the device.

[0107] S7: Use the trained neural network model. Given the features X device of the three-dimensional model of the station device base and the features X

[0108] such as Figure 2 shown, when the time step t is 1, the parameter (weight and bias) update process of the neural network model is as follows:

[0109] Assume the following coefficients are set: decay coefficient β1 = 0.9, β2 = 0.999, learning rate α = 0.001, small constant ∈ to prevent division by zero = 10 -8 ;

[0110] Initialize the first-order momentum m0 = 0, second-order momentum v0 = 0, and time step t = 0;

[0111] 1. Calculate the gradient g1 (time step t = 1, gradient is g t ):

[0112]

[0113] where θ is the current model parameter, which is the weight W or bias b;

[0114] 2. Update the momentum estimate:

[0115] m1 = 0.9·m0 + (1 - 0.9)·g1 = 0.1g1;

[0116] v1 = 0.999·v0 + (1 - 0.999)·g1 2 = 0.001g1 2 ;

[0117] 3. Bias correction:

[0118]

[0119] 4. Finally, update the model parameters:

[0120]

[0121] As the time step t increases, and will gradually approach 0. Therefore, the bias correction will tend to 1, making the momentum estimation more accurately reflect the true gradient information.

[0122] As Figure 3 shown, the operator first selects the device icon (taking the AFC turnstile as an example) from the 3D interface to issue a control command. The on-site device will perform corresponding actions, and the device will feedback the status to the 3D model device. At the same time, the model will have corresponding animated linkages, enhancing the operator's perception of the device status.

[0123] The beneficial technical effects of this embodiment are as follows:

[0124] (1) The integrated monitoring system can achieve multi-dimensional control and status display. It not only retains the functions of the integrated monitoring system for control and status monitoring on the 2D interface but also enables control on the visual 3D interface. The station equipment is represented by a series of English letters and numbers in the system. The operators no longer need to rely solely on experience and memory to operate the station. They can identify the equipment to be controlled based on the location of the equipment and the model icon, which is intuitive and efficient.

[0125] (2) The device model of the station is decoupled from the basic model. When the device configuration changes, it allows for the modification of device components. The model has high flexibility and scalability, can be customized and adjusted according to the changing requirements, and will not affect the unmodified model. The 3D model can be independently edited, so the generalizability of the model is strong. In the future, if new functions or technologies need to be integrated, it is not necessary to make large-scale modifications to the entire model, reducing the cost of model maintenance and improving the reusability.

[0126] (3) The animated linkages simulate the actions of real devices. When the operator executes a control command, they can see the response actions of the device on the model in real time. This real-time feedback not only enhances the operator's interactive experience but also helps the operator more quickly understand the effects of the control command and the responses of the device. It can more accurately monitor the operating status and actions of the device, reduce the risk of misoperation, and improve operation safety.

[0127] (4) The integrated monitoring system is not only applicable to daily operations, but also can be used for technical training and demonstration; the 3D model can more comprehensively display the position and dynamic changes of the equipment in space. Through the linkage of 3D vision and animation, new employees can more quickly understand, learn and master the equipment operation skills, reduce the training cost, and at the same time, it can also be used to demonstrate the technical capabilities and equipment functions to the outside world.

[0128] Although the above embodiments have detailed the concept and embodiments of the present invention with reference to the accompanying drawings, those of ordinary skill in the art can recognize that various improvements and transformations can still be made to the present invention without departing from the scope defined by the claims. Therefore, they will not be elaborated here one by one.

Claims

1. A BIM nervous system architecture for rail transit, characterized in that The BIM nervous system architecture includes a data and model combination module, and the data and model combination module includes a three-dimensional model of station equipment, a three-dimensional model of station foundation, and a neural network model. The three-dimensional model of station equipment is placed into the three-dimensional model of station foundation through a position matching algorithm, and the position matching algorithm is implemented through the neural network model. The neural network model is a multi-layer perceptron, and the multi-layer perceptron consists of an input layer, two hidden layers, and an output layer; The position matching algorithm includes the following steps: S1: Obtain the feature X of the 3D model of the station equipment device and the feature X of the 3D model of the station foundation base and perform preprocessing; S2: Calculate the position parameters of the device in the basic three-dimensional model of the station by using the neural network model S2.1: Input the feature X of the three-dimensional model of the station equipment device and the feature X of the three-dimensional model of the station foundation base into the input layer to calculate the input feature vector X; S2.2: Process the input feature vector X through the two hidden layers to obtain the output H1 of the first hidden layer and the output H2 of the second hidden layer in sequence; S2.3: Calculate the position parameters of the device in the 3D basic model of the station using the output layer S3: Train the neural network model using the loss function L; S4: Calculate the gradients of the loss function L with respect to the weights and biases of the neural network model; S5: Use the Adam optimization algorithm to update the weights and biases of the neural network model; S6: Repeat steps S2 to S5. Through multiple iterations, make the loss function L reach a convergence state, and the weights and biases of the neural network model are adjusted to the position that minimizes the loss function, so that the neural network model can accurately adjust the position parameters of the equipment; S7: Using the trained neural network model, given the feature X of the 3D model of the station equipment device and the feature X of the 3D model of the station foundation base , the neural network model outputs the position parameters of the equipment in the 3D model of the station foundation 2. The BIM nervous system architecture for rail transit according to claim 1, characterized in that In step S1 of the position matching algorithm, Feature X of the three-dimensional model of the station equipment device including the geometric shape, dimensions, and predefined spatial information of the equipment, and Feature X of the three-dimensional model of the station foundation base including the structure and spatial information of the building.

3. The BIM nervous system architecture for rail transit according to claim 2, characterized in that In step S2.1 of the position matching algorithm, The calculation formula for the input feature vector X is: X = [X device , X base ; where X is a vector of length d device + d base , d device is the dimension of the features of the three-dimensional model of the station equipment, and d base is the dimension of the features of the three-dimensional model of the station foundation.

4. The BIM nervous system architecture for rail transit according to claim 3, characterized in that In step S2.2 of the position matching algorithm, The calculation formula for the output H1 of the first hidden layer is: H1 = f(W1X + b1); where W1 is the weight matrix of the first hidden layer, with dimensions h1×(d device +d base ); b1 is the bias vector of the first hidden layer, with dimensions h1; f is the ReLU activation function; The calculation formula for the output H2 of the second hidden layer is: H2 = f(W2X + b2); In the formula, W2 is the weight matrix of the second hidden layer, with a dimension of h1×h2; b2 is the bias vector of the second hidden layer, with a dimension of h2.

5. The BIM nervous system architecture for rail transit according to claim 4, characterized in that In step S2.3 of the position matching algorithm, The position parameters of the device in the basic 3D model of the station The calculation formula is as follows: wherein, W3 is the weight matrix of the output layer, with a dimension of d output ×h2; b3 is the bias vector of the output layer, with a dimension of d output ; has a dimension of d output , if d output = 12, where 9 parameters represent the 9 elements of the rotation matrix R, the rotation matrix R is a 3×3 matrix, and 3 parameters represent the 3 elements of the translation vector T, the translation vector T is a 3×1 vector, including three components (x, y, z), respectively representing the translation distances along the x-axis, y-axis, and z-axis.

6. The BIM nervous system architecture for rail transit according to claim 5, characterized in that In step S3 of the position matching algorithm, The loss function L measures the difference between the predicted value and the true value Y. The loss function L uses the mean squared error: where n is the number of training samples, and Y i is the true position parameter of the i-th sample, and is the predicted position parameter of the i-th sample.

7. The BIM nervous system architecture for rail transit according to claim 6, characterized in that In step S4 of the position matching algorithm, The gradient of the loss function L with respect to the weight matrix W The calculation formula is as follows: The gradient of the loss function L with respect to the bias vector b is calculated as follows:

8. The BIM nervous system architecture for rail transit according to claim 7, characterized in that In step S5 of the position matching algorithm, Initialize the first-order momentum m and the second-order momentum v: where β1 and β2 are attenuation coefficients used to control the attenuation rate of historical information, usually set to 0.9 and 0.999; θ t is the weight / bias of the current model; Perform bias correction: Update the weights / biases of the neural network model: where ∈ is a small constant to prevent division by zero errors, usually set to 10 -8 ; α is the learning rate that controls the step size of weight / bias updates of the neural network model; t is the time step, tracking the number of current iterations. It increases by 1 for each iteration. and are both bias correction terms.

9. The BIM nervous system architecture for rail transit according to claim 1, characterized in that The BIM nervous system architecture further includes a control and execution module, and the control and execution module includes an intelligent control module and a real-time feedback module. The intelligent control module is used to control the operating state of the equipment, and the real-time feedback module is used to feedback the operating state of the equipment.

10. The BIM nervous system architecture for rail transit according to claim 1, characterized in that The BIM nervous system architecture further includes an animation linkage module, and the animation linkage module includes a keyframe animation module and a skeletal animation module. The keyframe animation module is used to simulate the basic actions of the equipment, and the skeletal animation module is used to simulate the movement of complex equipment.

Images