CN-122021336-A - Ethylene device hoisting operation control system and method based on BIM

CN122021336ACN 122021336 ACN122021336 ACN 122021336ACN-122021336-A

Abstract

The application relates to the technical field of pre-filled syringes, in particular to the technical field of packaging strategy generation, and in particular relates to a BIM (building information modeling) space simulation module for digitally modeling an operation area, a hoisting target and a hoisting position to generate a BIM virtual operation space; the system comprises a hoisting path generation module, a hoisting strategy generation module and a hoisting strategy generation module, wherein the hoisting strategy generation module is used for carrying out hoisting state decomposition and screening on path elements in the alternative path set Cr by utilizing a neural network and outputting an optimal hoisting path and a hoisting state sequence, and the standardized data set Ds of the BIM virtual operation space is obtained to generate the alternative path set Cr. The closed loop hoisting strategy generation is completed, the effective strategy generation of the whole hoisting operation is realized, the hoisting efficiency is improved, and the operation safety is enhanced.

Inventors

YU HAOTIAN
ZHANG QINCHAO
WANG NANA
WU YANTONG
SUN YING
LI ZEBIN
CHANG LI

Assignees

山东海湾吊装工程股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260211

Claims (10)

1. The ethylene device hoisting operation control system based on BIM is characterized by comprising a BIM space simulation module, a hoisting path generation module and a hoisting strategy generation module, wherein the functions of the modules are as follows: The BIM space simulation module is used for digitally modeling the operation area, the hoisting target and the hoisting position to generate a BIM virtual operation space; The hoisting path generation module is used for acquiring a standardized data set Ds of the BIM virtual operation space and generating an alternative path set Cr; the hoisting strategy generation module is used for decomposing and screening hoisting states of path elements in the alternative path set Cr by utilizing the neural network, and outputting an optimal hoisting path and a hoisting state sequence; the hoisting state is the instantaneous space pose of the hoisting target at a certain moment in the hoisting process.
2. The BIM-based ethylene plant hoisting operation control system of claim 1, wherein: the normalized data set Ds includes collision detection information, which is acquired based on the separation axis theorem SAT.
3. The BIM-based ethylene plant hoisting operation control system of claim 2, wherein: The hoisting path generation module receives the standardized data set Ds and generates an alternative path set Cr by using a modified type a path generation algorithm, and the method comprises the following steps: loading differential preset constraint based on hoisting target classification; Secondly, designing an improved cost function, and introducing a contour safety distance penalty term and a gesture change penalty term: In the formula, As the total cost of the current node n, For the accumulated three-dimensional path length from the start point S to the current node n, For the heuristic distance of node n to endpoint T, For the attitude change penalty term, the cumulative attitude angle change amount from the start point S to the current node n is associated, When the minimum angle change from the last node to the node n is related to the lifting target gesture, the nearest distance between the surface of the lifting target profile and the obstacle is used as the profile safety distance punishment item, 、、 Different values are set for punishment item weight coefficients based on different hoisting types; When the search loop is executed, only the paths are saved without stopping the loop after the end point T is found, the rest nodes are processed OpenList continuously, all possible paths are obtained, a preset number of alternative paths are obtained through differential preset constraint screening and total cost minimum principle, and an alternative path set Cr is generated.
4. The BIM-based ethylene plant hoisting operation control system of claim 1, wherein: the hoisting strategy generation module is a neural network model based on a CNN-transducer fusion framework, takes a standardized data set Ds of a BIM space simulation module, an alternative path set Cr of a hoisting path generation module and environmental parameters of an operation area as inputs, directly outputs an optimal hoisting state sequence and an optimal hoisting path from end to end, and comprises an input layer, a path-node coding layer, an attention layer, a double-branch characteristic fusion layer and a state sequence generation layer, wherein the functions of the layers are as follows: an input layer for normalizing the input characteristic data, performing characteristic splicing according to the dimension of the path node, outputting an original characteristic set, (I, j) represents a j-th node of the i-th path; the path-node coding layer adds path attribution identification for each node and outputs node characteristics carrying path attribution information ; An attention layer for extracting the space and time sequence characteristics of nodes in the same path, shielding the characteristic interaction of the cross-path nodes by an attention mask, and outputting a global node characteristic set ; The branch feature fusion layer extracts space matching features between nodes and hoisting states through CNN branches, extracts time sequence optimization features of state sequences through a transducer branch, and fuses the two branch features to output fusion feature vectors ; And a state sequence generation layer, mapping the fusion characteristics into an attitude angle sequence, enabling each state to be attached to a node of the same path, and outputting an optimal state sequence and an optimal path.
5. The BIM-based ethylene plant hoisting operation control system of claim 4, wherein: The input layer firstly eliminates dimension and standardization of each characteristic parameter, a standardization algorithm uses a Z-Score standardization algorithm to splice standardized characteristics according to path node dimension to obtain an original characteristic set 。
6. The BIM-based ethylene plant hoisting operation control system of claim 4, wherein: The path-node coding layer adds path attribution identification for each node through independent thermal coding and feature fusion, and outputs node features carrying path attribution information The mathematical expression is as follows: In the formula, For the one-hot encoding vector of the ith path, the dimension is the number of alternative paths, emb is a learnable full-connection layer for mapping the low-dimensional one-hot encoding to a high-dimensional home embedded vector, Embedding vector for the ith path, unique identifier for the path, the tie is an element-by-element addition operation, Representing j nodes on the i-th alternative path.
7. The BIM-based ethylene plant hoisting operation control system of claim 4, wherein: the attention layer comprises a path space coding sub-layer and a perception attention sub-layer; The path space coding sublayer carries out 1D-CNN coding on node characteristics in a single path, extracts space continuous characteristics of nodes in the path, independently codes different paths, and outputs coding characteristic sets { of all paths }; The awareness sublayer to normalize BIM features For inquiring vectors, taking the coding features of the same path CNN as key/value vectors, shielding the cross-path attention interaction through a path attribution attention mask MID, and finally outputting a global node feature set The mathematical expression is as follows: where Q is the query vector, For the key/value vector of the i-th path, For the single-head dimension of multi-head attention, the path attribution attention mask MID takes the value rule that if the node for calculating the attention and the matching node of the query vector take the value of being crossing paths Infinity, the value of the same path is 0, Is the home perceived attention feature of the ith path.
8. The BIM-based ethylene plant hoisting operation control system of claim 4, wherein: The CNN branches in the branch feature fusion layer calculate candidate attitude angles meeting collision constraint for each node, and ensure that the contour of the equipment at the node is matched with the space constraint; The transducer branch models switching logic of adjacent states, and autonomously learns sequential logic characteristics; The branch feature fusion layer is provided with a gating feature fusion mechanism, so that the model automatically adjusts the contribution degree of the space matching feature and the time sequence optimizing feature, and the mathematical expression is as follows: In the formula, And To gate the learnable weight matrix and bias, sigmoid activates the function, Representing the resulting output of the fused feature vector, And The output feature vectors of the two branches, respectively.
9. The BIM-based ethylene plant hoisting operation control system of claim 4, wherein: the state sequence generation layer firstly maps the fusion characteristics into an attitude angle sequence matched with the task length through 2 layers of full-connection layers, and the state sequence generation layer specifically comprises the following steps: where, reLU represents the activation function, 、 Is an offset vector corresponding to two full-connection layers; Secondly, performing spatial similarity matching on the generated attitude angle sequence and nodes of the all-alternative path to obtain a primary home path corresponding to the state sequence, wherein the mathematical expression is as follows: In the formula, And the node coordinates representing the binding of the t-th attitude angle are used for binding the attitude angle with the nodes in the same path through the attribution relation of the nodes and the paths and the sequence of the path nodes.
10. The ethylene device hoisting operation control method based on BIM is characterized by comprising the following steps: s100, a BIM space simulation module, which is used for digitally modeling an operation area, a hoisting target and a hoisting position to generate a BIM virtual operation space; S200, a hoisting path generation module acquires a standardized data set Ds of the BIM virtual operation space and generates an alternative path set Cr; and S300, a hoisting strategy generation module, namely performing hoisting state decomposition and screening on path elements in the alternative path set Cr by utilizing a neural network, and outputting an optimal hoisting path and a hoisting state sequence.

Description

Ethylene device hoisting operation control system and method based on BIM Technical Field The application relates to the technical field of hoisting strategy generation, in particular to a control system and a control method for hoisting operation of an ethylene device based on BIM. Background The ethylene device is core equipment in petrochemical industry, lifting operation of the ethylene device relates to heavy equipment such as a large-scale tower, a reactor, a heat exchanger and the like, the weight of a single equipment can reach hundreds of tons, the installation height is hundreds of meters, and an operation scene has the remarkable characteristics of complex space constraint, dense barriers, severe process requirements and the like. In the hoisting process, the attitude control and path planning of the equipment directly determine the safety, efficiency and economy of the operation, and are key bottleneck links for the construction of ethylene devices. For a long time, attitude control planning of hoisting operation of an ethylene device mainly depends on manual experience driving, and is assisted by basic space simulation of BIM (building information model) to carry out simple path planning, so that the intelligent degree of overall operation planning is low, the effectiveness of the path planning is insufficient and is disjointed with attitude control, and the overall operation efficiency and safety are insufficient. Disclosure of Invention The application provides a control system and a control method for hoisting operation of an ethylene device based on BIM, which are used for solving or partially solving the problems in the background technology. The application provides a control system for hoisting operation of an ethylene device based on BIM, which comprises a BIM space simulation module, a hoisting path generation module and a hoisting strategy generation module, wherein the functions of the modules are as follows: The BIM space simulation module is used for digitally modeling the operation area, the hoisting target and the hoisting position to generate a BIM virtual operation space; The hoisting path generation module is used for acquiring a standardized data set Ds of the BIM virtual operation space and generating an alternative path set Cr; the hoisting strategy generation module is used for decomposing and screening hoisting states of path elements in the alternative path set Cr by utilizing the neural network, and outputting an optimal hoisting path and a hoisting state sequence; the hoisting state is the instantaneous space pose of the hoisting target at a certain moment in the hoisting process. Preferably, the normalized data set Ds includes collision detection information, which is acquired based on the separation axis theorem SAT. Preferably, the hoisting path generation module receives the normalized data set Ds and generates the alternative path set Cr by using a modified a-path generation algorithm, and includes the following steps: loading differential preset constraint based on hoisting target classification; Secondly, designing an improved cost function, and introducing a contour safety distance penalty term and a gesture change penalty term: In the formula, As the total cost of the current node n,For the accumulated three-dimensional path length from the start point S to the current node n,For the heuristic distance of node n to endpoint T,For the attitude change penalty term, the cumulative attitude angle change amount from the start point S to the current node n is associated,When the minimum angle change from the last node to the node n is related to the lifting target gesture, the nearest distance between the surface of the lifting target profile and the obstacle is used as the profile safety distance punishment item,、、Different values are set for punishment item weight coefficients based on different hoisting types; When the search loop is executed, only the paths are saved without stopping the loop after the end point T is found, the rest nodes are processed OpenList continuously, all possible paths are obtained, a preset number of alternative paths are obtained through differential preset constraint screening and total cost minimum principle, and an alternative path set Cr is generated. Preferably, the hoisting strategy generation module is a neural network model based on a CNN-Transformer fusion architecture, takes a standardized data set Ds of a BIM space simulation module, an alternative path set Cr of a hoisting path generation module and environmental parameters of an operation area as inputs, and directly outputs an optimal hoisting state sequence and an optimal hoisting path from end to end, wherein the method comprises an input layer, a path-node coding layer, an attention layer, a double-branch feature fusion layer and a state sequence generation layer, and the functions of the layers are as follows: an input layer for normalizing the input characteristic da