Data simulation method and system based on open-pit mine 5g network

Abstract

The application provides a data simulation method and system based on an open-pit mine 5G network, and relates to the technical field of artificial intelligence.In the application, two network topology data are acquired;first and second topology data segment sets of the two network topology data are constructed;feature vectors of the first and second network topology data are determined based on the first and second topology data segment sets;new network topology data are generated based on the aggregation result of the feature vectors of the first and second network topology data;when the simulation result of the new network topology data meets a preset simulation requirement, the simulation is determined to be ended, and the new network topology data are determined as target network topology data.Based on the above method, the reliability of the network topology data determined in the data simulation process can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a data simulation method and system based on a 5G network in an open-pit mine. Background Art

[0002] Open-pit mines can use 5G networks for communication and data transmission. 5G networks offer high speed, low latency, and large capacity, making efficient wireless communication possible in open-pit mines. Through 5G networks, miners can monitor equipment and data in real time, perform remote operations and make adjustments, and improve production efficiency and safety. Furthermore, 5G networks can support the connection of IoT devices, enabling intelligent management and automated control, further optimizing mine operations. In summary, 5G networks are a crucial technological enabler for open-pit mines and offer numerous benefits. Existing technologies typically determine network performance indicators, such as capacity, throughput, latency, and connectivity. Based on these performance indicators, a corresponding network layout, or network topology data (including base station locations, antenna orientations, and transmission links), is generated. Modeling and simulation are then performed based on this network topology data to generate simulation results. If the simulation results match the previous network performance indicators, the final network topology data is obtained and simulation is discontinued. If the simulation results do not match the previous network performance indicators, the network topology data needs to be adjusted and simulation repeated.

[0003] In the prior art, in order to improve the reliability of network layout determination, multiple network topology data are generally generated for separate simulations. When the simulation results of multiple (two) network topology data meet the performance index requirements of the network, one network topology data can be selected from them as the final network topology data. However, if the selection is based on arbitrariness (different network topology data may have different focuses, such as one focusing on small engineering volume and the other focusing on low equipment cost), it is easy to cause the reliability of the final network topology data to be relatively low. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a data simulation method and system based on the open-pit mine 5G network, so that the reliability of the network topology data determined during the data simulation process can be improved.

[0005] To achieve the above objectives, the embodiments of the present invention adopt the following technical solutions:

[0006] A data simulation method based on an open-pit mine 5G network comprises: obtaining two network topology data whose simulation results for the open-pit mine 5G network meet preset simulation requirements, wherein the network topology data is used to characterize the network layout in the open-pit mine 5G network, and the simulation results are used to characterize the performance indicators of the open-pit mine 5G network; constructing a first topology data segment set of the first network topology data of the two network topology data and a second topology data segment set of the second network topology data of the two network topology data, wherein each first topology data segment in the first topology data segment set belongs to the first network topology data, and each second topology data segment in the second topology data segment set belongs to the first network topology data. The topology data fragment belongs to the second network topology data; using the target topology data to analyze the network, based on the first topology data fragment set of the first network topology data and the second topology data fragment set of the second network topology data, determine the characteristic vector of the first network topology data and the characteristic vector of the second network topology data; using the target topology data to analyze the network, based on the aggregation result of the characteristic vector of the first network topology data and the characteristic vector of the second network topology data, generate new network topology data, and when the simulation result of the new network topology data meets the preset simulation requirements, determine that the simulation ends, and use the new network topology data as the target network topology data.

[0007] In some preferred embodiments, in the above-mentioned data simulation method based on the open-pit mine 5G network, the step of using the target topology data analysis network to determine the feature vector of the first network topology data and the feature vector of the second network topology data based on the first topology data fragment set of the first network topology data and the second topology data fragment set of the second network topology data includes: loading the first topology data fragment set into the target topology data analysis network, and using the shared focusing unit and the first feature mining unit in the target topology data analysis network to mine the corresponding first candidate topology vector; loading the second topology data fragment set into the target topology data analysis network, and using the shared focusing unit and the second feature mining unit in the target topology data analysis network to mine the corresponding first candidate topology vector. The first candidate topology vector and the second candidate topology vector are loaded into the shared focusing unit and the joint feature mining unit included in the target topology data analysis network, and the focused feature mining results of the first candidate topology vector and the focused feature mining results of the second candidate topology vector are aggregated by using the joint feature mining unit to form a corresponding aggregated feature vector, and the aggregated feature vector includes the feature vector of the first network topology data and the feature vector of the second network topology data; wherein the shared focusing unit is used to perform focused feature mining on the loaded data, and the first feature mining unit, the second feature mining unit and the joint feature mining unit are respectively used to mine the corresponding focused feature mining results to form a corresponding feature vector.

[0008] In some preferred embodiments, in the above-mentioned data simulation method based on the open-pit mine 5G network, the data simulation method based on the open-pit mine 5G network further includes: determining the training first data of the training first network topology data, the training first data is used to indicate the training first topology data segment set, and the training first topology data segment set includes multiple training first topology data segments in the training first network topology data; loading the training first data into the target teacher model and outputting the corresponding first feature vector; overwriting at least one training first topology data segment in the training first topology data segment set indicated by the training first data to form corresponding training second data; loading the training second data into the target topology model. In the data analysis network, the corresponding second eigenvector is output. The target topology data analysis network belongs to the target model relative to the target teacher model in the model migration, and improves its own performance by transferring knowledge from the target teacher model. The target topology data analysis network is used to mine the eigenvector; based on the first eigenvector and the second eigenvector, a first error index is calculated, and the first error index is used to reflect the difference between the first eigenvector and the second eigenvector; based on the first error index, the target teacher model and the target topology data analysis network are updated and optimized to form an updated target teacher model and an updated target topology data analysis network.

[0009] In some preferred embodiments, in the above-mentioned data simulation method based on the open-pit mine 5G network, the training first data includes a training first topology data segment set and a topology data segment fusion data, and the step of determining the training first data of the training first network topology data includes: performing data perturbation on the training first network topology data to form the perturbed first network topology data of the training first network topology data; performing data segmentation on the perturbed first network topology data to form the training first topology data segment set of the perturbed first network topology data; determining the topology data segment fusion data of the perturbed first network topology data based on the training first topology data segment set of the perturbed first network topology data to form the topology data segment fusion data of the perturbed first network topology data, the training first topology data segment set includes the corresponding embedding vectors of multiple training first topology data segments, and the topology data segment fusion data is obtained by fusing the embedding vectors of the multiple training first topology data segments.

[0010] In some preferred embodiments, in the above-mentioned data simulation method based on the open-pit mine 5G network, the first feature vector includes a first topological data vector corresponding to the training first topological data segment set and a second topological data vector corresponding to the topological data segment fusion data, and the second feature vector includes a third topological data vector corresponding to the training first topological data segment set and a fourth topological data vector corresponding to the topological data segment fusion data. The step of calculating the first error index based on the first feature vector and the second feature vector includes: calculating a first local error index based on the first topological data vector and the third topological data vector, and the first local error index is used to reflect the difference between the first topological data vector and the third topological data vector; calculating a second local error index based on the second topological data vector and the fourth topological data vector, and the second local error index is used to reflect the difference between the second topological data vector and the fourth topological data vector; calculating the first error index based on the first local error index and the second local error index.

[0011] In some preferred embodiments, in the above-mentioned data simulation method based on the open-pit mine 5G network, the step of calculating the first local error index based on the first topological data vector and the third topological data vector includes: based on the set coordinates of the at least one covered training first topological data segment in the training first topological data segment set, extracting the local first topological data vector corresponding to the set coordinates in the first topological data vector, and extracting the local third topological data vector corresponding to the set coordinates in the third topological data vector; calculating the corresponding first local error index based on the local first topological data vector and the local third topological data vector.

[0012] In some preferred embodiments, in the above-mentioned data simulation method based on the open-pit mine 5G network, the target topology data analysis network includes a shared focusing unit, a first feature mining unit, a second feature mining unit and a joint feature mining unit, the shared focusing unit is used to perform focused feature mining on the loaded data, the first feature mining unit, the second feature mining unit and the joint feature mining unit are respectively used to mine the corresponding focused feature mining results to form corresponding feature vectors, the data simulation method based on the open-pit mine 5G network also includes: using the topological data segmentation model to segment the training second network topology data to form a corresponding third feature vector, the third feature vector includes the feature vectors of the multiple training second topology data segments formed by segmenting the training second network topology data; at least one of the training second topology data segments corresponding to the third feature vector is segmented. The invention covers the at least one training second topology data segment to form a corresponding fourth eigenvector; the fourth eigenvector is loaded into the target topology data analysis network to mine the corresponding fifth eigenvector; based on the set coordinates of the at least one covered training second topology data segment in the training second topology data segment set, the local second topology data vector corresponding to the set coordinates in the fifth eigenvector is extracted; based on the local second topology data vector and the local fourth topology data vector, a second error index is calculated, the local fourth topology data vector is the eigenvector of the at least one training second topology data segment included in the fourth eigenvector, and the second error index is used to reflect the difference between the local second topology data vector and the local fourth topology data vector; based on the second error index, the second feature mining unit in the target topology data analysis network is updated and optimized.

[0013] In some preferred embodiments, in the above-mentioned data simulation method based on the open-pit mine 5G network, the target topology data analysis network includes a shared focusing unit, a first feature mining unit, a second feature mining unit and a joint feature mining unit. The shared focusing unit is used to perform focused feature mining on the loaded data, and the first feature mining unit, the second feature mining unit and the joint feature mining unit are respectively used to mine the corresponding focused feature mining results to form corresponding feature vectors. The data simulation method based on the open-pit mine 5G network also includes: determining a training network topology data combination, the training network topology data combination including two network topology data whose corresponding simulation results meet the preset simulation requirements; using the target teacher model to determine the sixth feature vector, and using the topology data segmentation model to determine the seventh feature vector, the sixth feature vector is the feature vector of the first training network topology data in the training network topology data combination, and the seventh feature vector is the feature vector of the second training network topology data in the training network topology data combination. the characteristic vector of the training data; covering at least one training first topology data segment in the training first topology data segment set corresponding to the sixth characteristic vector, and mining out the corresponding eighth characteristic vector; covering at least one training second topology data segment in the training second topology data segment set corresponding to the seventh characteristic vector, and mining out the corresponding ninth characteristic vector; loading the eighth characteristic vector and the ninth characteristic vector into the target topology data analysis network respectively to form a training aggregate characteristic vector corresponding to the training network topology data combination; calculating a third error index based on the training aggregate characteristic vector, the sixth characteristic vector and the seventh characteristic vector, the third error index being used to reflect the difference between the characteristic vector of the first training network topology data in the training aggregate characteristic vector and the sixth characteristic vector, and the difference between the characteristic vector of the second training network topology data in the training aggregate characteristic vector and the seventh characteristic vector; updating and optimizing the target topology data analysis network based on the third error index.

[0014] In some preferred embodiments, in the above-mentioned data simulation method based on the open-pit mine 5G network, the target topology data analysis network includes multiple network hierarchical structures, the first network hierarchical structure in the multiple network hierarchical structures includes a shared focusing unit, a second feature mining unit and a first feature mining unit, and the second network hierarchical structure in the multiple network hierarchical structures includes a shared focusing unit, a second feature mining unit, a first feature mining unit and a joint feature mining unit; the steps of loading the eighth eigenvector and the ninth eigenvector into the target topology data analysis network respectively to form the training aggregated eigenvector corresponding to the training network topology data combination include: loading the eighth eigenvector into the first network hierarchical structure in sequence; The shared focusing unit and the first feature mining unit output the corresponding training first candidate topology vector; the ninth feature vector is sequentially loaded into the shared focusing unit and the second feature mining unit in the first network hierarchical structure, and the corresponding training second candidate topology vector is output; the training first candidate topology vector is sequentially loaded into the shared focusing unit and the joint feature mining unit in the second network hierarchical structure, and the training second candidate topology vector is sequentially loaded into the shared focusing unit and the joint feature mining unit in the second network hierarchical structure, and the training first candidate topology vector and the training second candidate topology vector are aggregated by using the joint feature mining unit to form a training aggregated feature vector corresponding to the training network topology data combination.

[0015] An embodiment of the present invention also provides a data simulation system based on an open-pit mine 5G network, comprising a processor and a memory, wherein the memory is used to store a computer program, and the processor is used to execute the computer program to implement the above-mentioned data simulation method based on an open-pit mine 5G network.

[0016] The data simulation method and system based on the open-pit mine 5G network provided by the embodiment of the present invention can obtain two network topology data; construct a first topology data fragment set and a second topology data fragment set of the two network topology data; based on the first topology data fragment set and the second topology data fragment set, determine the characteristic vector of the first network topology data and the characteristic vector of the second network topology data; based on the aggregation result of the characteristic vector of the first network topology data and the characteristic vector of the second network topology data, generate new network topology data, and when the simulation result of the new network topology data meets the preset simulation requirements, determine that the simulation is over and use the new network topology data as the target network topology data. Based on the foregoing content, after obtaining the two network topology data whose simulation results of the network simulation for the open-pit mine 5G network meet the preset simulation requirements, it is not simply to determine one network topology data from the two network topology data, but to mine the two topology data, and then fuse the mined feature vectors, so that new network topology data can be generated based on the fusion results. In this way, the new network topology data can take into account the two obtained network topology data. Therefore, when the simulation results of the new network topology data meet the preset simulation requirements, the reliability of the new network topology data as the target network topology data can be guaranteed, so that the reliability of the network topology data determined during the data simulation process can be improved, thereby improving the problem of relatively low reliability in the existing technology.

[0017] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A structural block diagram of a data simulation system based on a 5G network for open-pit mines provided by an embodiment of the present invention;

[0019] Figure 2 A schematic flow chart of the steps included in the data simulation method based on the 5G network of an open-pit mine provided in an embodiment of the present invention;

[0020] Figure 3 A schematic diagram of the modules included in the data simulation device based on the open-pit mine 5G network provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0022] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0023] like Figure 1 As shown, an embodiment of the present invention provides a data simulation system based on an open-pit mine 5G network. The data simulation system based on an open-pit mine 5G network may include a memory and a processor.

[0024] In detail, in one embodiment, the memory and the processor are electrically connected directly or indirectly to achieve data transmission or interaction. For example, they can be electrically connected to each other through one or more communication buses or signal lines. The memory can store at least one software function module (computer program) that can exist in the form of software or firmware. The processor can be used to execute the executable computer program stored in the memory, thereby implementing the data simulation method based on the open-pit mine 5G network provided by the embodiment of the present invention.

[0025] Specifically, in one embodiment, the memory may be, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), etc. The processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a system on chip (SoC), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0026] In detail, in one embodiment, the data simulation system based on the open-pit mine 5G network can be a server with data processing capabilities.

[0027] Combine Figure 2 The present invention also provides a data simulation method based on a 5G network in an open-pit mine, which can be applied to the above-mentioned data simulation system based on a 5G network in an open-pit mine. The method steps defined in the process related to the data simulation method based on a 5G network in an open-pit mine can be implemented by the data simulation system based on a 5G network in an open-pit mine.

[0028] The following will Figure 2 The specific process shown is explained in detail.

[0029] Step S110: obtaining two network topology data whose simulation results for the open-pit mine 5G network meet the preset simulation requirements.

[0030] In an embodiment of the present invention, a data simulation system based on an open-pit mine 5G network can obtain two network topology data whose simulation results for the network simulation of the open-pit mine 5G network meet the preset simulation requirements. The two network topology data can be constructed based on different network layout teams respectively, or can be constructed based on a network layout team based on different biases (irrelevant to network performance indicators, such as engineering volume, equipment cost, construction difficulty, etc.), and after simulation, the simulation results obtained meet the preset simulation requirements (preset performance indicators). The network topology data is used to characterize the network layout in the open-pit mine 5G network, such as the location of the base station, the direction of the antenna, the transmission link, etc., and the simulation results are used to characterize the performance indicators of the open-pit mine 5G network, such as network capacity, throughput, delay, connectivity, etc. In other words, the network topology data can be modeled first, and then, using simulation tools such as NS-3, OMNeT++, etc. (the specific simulation process is not limited and can refer to relevant existing technologies), a suitable protocol model is selected, and the generated traffic is injected into the simulation scenario, which can simulate the communication process between devices, including transmission, processing delay, interference, etc. The performance indicators of the network are analyzed through simulation results, and then compared with the preset performance indicators to determine whether the preset simulation requirements are met.

[0031] Step S120 : constructing a first topology data segment set of a first network topology data of the two network topology data and a second topology data segment set of a second network topology data of the two network topology data.

[0032] In an embodiment of the present invention, a data simulation system based on an open-pit mine 5G network can construct a first topology data fragment set of the first network topology data in the two network topology data and a second topology data fragment set of the second network topology data in the two network topology data. Each first topology data fragment in the first topology data fragment set belongs to the first network topology data, and each second topology data fragment in the second topology data fragment set belongs to the second network topology data. Among them, the first topology data fragments can be arranged in sequence so that the first topology data fragment set is a valid set, and the second topology data fragments can be arranged in sequence so that the second topology data fragment set is a valid set. When sorting in sequence, they can be sorted according to their order in the corresponding network topology data. The network topology data can be a text data or data including text and drawings.

[0033] Step S130 , analyzing the network using the target topology data, and determining a feature vector of the first network topology data and a feature vector of the second network topology data based on a first topology data segment set of the first network topology data and a second topology data segment set of the second network topology data.

[0034] In an embodiment of the present invention, a data simulation system based on an open-pit mine 5G network can utilize the target topology data to analyze the network, and based on a first topology data fragment set of the first network topology data and a second topology data fragment set of the second network topology data, determine the feature vector of the first network topology data and the feature vector of the second network topology data. For example, the first topology data fragment set can be mined to obtain the feature vector of the first network topology data, and the second topology data fragment set can be mined to obtain the feature vector of the second network topology data. The feature vector is used to reflect the semantics of the corresponding network topology data. Among them, the granularity of the topology data fragments in the first topology data fragment set and the second topology data fragment set is not restricted. For example, for text data, it can be a granularity such as a sentence or paragraph, and for drawings, it can be a single drawing as the granularity. The specific selection can be made according to actual needs.

[0035] Step S140, using the target topology data to analyze the network, generating new network topology data based on the aggregation results of the feature vector of the first network topology data and the feature vector of the second network topology data, and when the simulation results of the new network topology data meet the preset simulation requirements, determining that the simulation is over, and using the new network topology data as the target network topology data.

[0036] In an embodiment of the present invention, a data simulation system based on an open-pit mine 5G network can utilize a target topology data analysis network to generate new network topology data based on the aggregation result of the feature vector of the first network topology data and the feature vector of the second network topology data, and when the simulation result of the new network topology data meets the preset simulation requirements (i.e., modeling is performed based on the new network topology data, and then simulation and other processing are performed), the simulation is determined to be completed, and the new network topology data is used as the target network topology data. In this way, the target network topology data can be used as the final layout of the open-pit mine 5G network, and then the construction of the open-pit mine 5G network can be carried out based on the target network topology data. Among them, the target topology data analysis network can be a convolutional neural network (CNN), which can include two parts, an encoding subnetwork (Encoder) and a decoding subnetwork (Decoder). The encoding subnetwork can be used to perform feature encoding to obtain corresponding feature vectors, such as executing step S130, and the decoding subnetwork can be used to perform feature decoding, such as executing step S140 to generate (or reconstruct) new network topology data.

[0037] Based on the foregoing content, after obtaining the two network topology data whose simulation results of the network simulation for the open-pit mine 5G network meet the preset simulation requirements, it is not simply to determine one network topology data from the two network topology data, but to mine the two topology data, and then fuse the mined feature vectors, so that new network topology data can be generated based on the fusion results. In this way, the new network topology data can take into account the two obtained network topology data. Therefore, when the simulation results of the new network topology data meet the preset simulation requirements, the reliability of the new network topology data as the target network topology data can be guaranteed, so that the reliability of the network topology data determined during the data simulation process can be improved, thereby improving the problem of relatively low reliability in the existing technology.

[0038] For example:

[0039] The first network topology data, that is, topology data A, can be:

[0040] Assume that topology data A represents a 5G network in an open-pit mine with five base stations. Each base station has three antennas to provide network coverage and communication connectivity. These base stations are interconnected via fiber optic links. In topology data A, base station A is located at (10, 20) with an antenna facing west; base station B is located at (30, 15) with an antenna facing east; base station C is located at (25, 35) with an antenna facing south; base station D is located at (5, 40) with an antenna facing north; and base station E is located at (20, 25) with an antenna facing west.

[0041] The second network topology data, namely topology data B, may be:

[0042] Assume that topology data B represents a 5G network in an open-pit mine with six base stations. Each base station has two antennas and uses a more economical wireless transmission method. These base stations communicate with each other via microwave links. In topology data B, base station F is located at (12, 18) with an antenna facing west; base station G is located at (28, 22) with an antenna facing east; base station H is located at (32, 30) with an antenna facing south; base station I is located at (15, 38) with an antenna facing north; base station J is located at (22, 26) with an antenna facing west; and base station K is located at (40, 20) with an antenna facing east.

[0043] Comparative analysis: Although topology data A and topology data B both meet the same network performance indicator requirements, they differ in terms of construction workload and equipment costs. Topology data A requires less construction workload because it has fewer base stations and uses fiber optic links, which may require less physical wiring and civil engineering work, thereby reducing the construction workload. Topology data B has a smaller equipment cost because it uses fewer antennas and an economical wireless transmission method, which may reduce equipment procurement and deployment costs, thereby reducing overall equipment costs.

[0044] In addition, both of the above network topology data can meet the following performance requirements:

[0045] Network capacity: Each base station is required to support simultaneous connection of 500 devices and provide stable data transmission. Each base station should have sufficient bandwidth to meet the needs of each connected user and ensure that the network is not congested due to too many connections. Throughput: Each base station is required to provide a total throughput of at least 10Gbps to meet the needs of large-scale data transmission and high-bandwidth applications. When tasks such as high-definition video transmission and large file transfer are carried out simultaneously in the mine, the network can provide sufficient bandwidth to ensure fast and smooth data transmission. Latency: Network latency is required to be kept below 5 milliseconds to support real-time control and monitoring applications. When equipment needs to be remotely controlled or monitored in real time in the mine, a low-latency network can ensure that the delay between operation and feedback is small, improving work efficiency and safety. Connectivity: The network is required to have wide coverage and good signal strength to ensure that devices throughout the mine can be stably connected to the network. For example, different areas in the mine, including mining areas and transportation areas, need to be covered by base stations, and the signal strength must be strong enough to ensure a stable connection.

[0046] It should be noted that the above content is only a simplified example. In actual applications, network topology data will be more complex and the data volume will be larger.

[0047] Specifically, in one embodiment, the above step S130 may include:

[0048] The first topological data fragment set is loaded into the target topological data analysis network, and the corresponding first candidate topological vector is mined using the shared focusing unit and the first feature mining unit in the target topological data analysis network. That is, the first topological data fragment set (the corresponding embedding vector, that is, the first topological data fragment set can be embedded to obtain the corresponding embedding vector) is first processed using the shared focusing unit, and then the processed data is processed by the first feature mining unit to obtain the first candidate topological vector;

[0049] The second topological data fragment set is loaded into the target topological data analysis network, and the corresponding second candidate topological vector is mined using the shared focusing unit and the second feature mining unit in the target topological data analysis network. That is, the second topological data fragment set (corresponding embedding vector) is first processed using the shared focusing unit, and then the processed data is processed by the second feature mining unit to obtain the second candidate topological vector;

[0050] The first candidate topology vector and the second candidate topology vector are loaded into the shared focusing unit and the joint feature mining unit included in the target topology data analysis network. The focused feature mining results of the first candidate topology vector and the focused feature mining results of the second candidate topology vector are aggregated by using the joint feature mining unit to form a corresponding aggregated feature vector. The aggregated feature vector includes the feature vector of the first network topology data and the feature vector of the second network topology data. For example, the first candidate topology vector can be loaded into the shared focusing unit and the joint feature mining unit in the second network hierarchical structure for processing in turn to obtain a first vector. The second candidate topology vector can be loaded into the shared focusing unit and the joint feature mining unit in the second network hierarchical structure for processing in turn to obtain a second vector. The first vector and the second vector are aggregated by using the joint feature mining unit, such as splicing (cascading), to form an aggregated feature vector corresponding to the network topology data combination. The shared focusing unit is used to perform focused feature mining on the loaded data. The first feature mining unit, the second feature mining unit and the joint feature mining unit are respectively used to mine the corresponding focused feature mining results to form a corresponding feature vector. The shared focusing unit can be shared by the first feature mining unit, the second feature mining unit, and the joint feature mining unit, that is, connected to the first feature mining unit, the second feature mining unit, and the joint feature mining unit respectively. The shared focusing unit can be a self-attention model. The first feature mining unit, the second feature mining unit, and the joint feature mining unit can all be feedforward propagation neural networks. The specific network parameters can be different and are formed during the corresponding network optimization process. They can belong to the aforementioned encoding subnetwork.

[0051] In detail, in one embodiment, in order to enable the target topology data analysis network to effectively perform the above step S130, the data simulation method based on the open-pit mine 5G network may further include the following network update process:

[0052] Determining training first data for training first network topology data, where the training first data is used to indicate a training first topology data segment set, where the training first topology data segment set includes a plurality of training first topology data segments in the training first network topology data, and the plurality of training first topology data segments can be arranged in sequence to form an ordered set;

[0053] Load the first training data into the target teacher model and output the corresponding first eigenvector. In deep learning, a teacher model refers to a model or network used to guide training. It is usually designed to have high performance and accuracy, and as a "teacher", it can provide the student model with information about the correct answer or target value. The main goal of the teacher model is to guide the training process of the student model through its output to improve the performance of the student model. This training method is usually called knowledge transfer. The teacher model can be a complex deep neural network model, or an integrated model or pre-trained model. It can achieve high accuracy by training on a large amount of data, and then transfer its knowledge to the student model.

[0054] Covering (i.e., masking) at least one training first topology data segment in the training first topology data segment set indicated by the training first data to form corresponding training second data; that is, the at least one training first topology data segment can be hidden;

[0055] The second training data is loaded into the target topology data analysis network, and the corresponding second eigenvector is output. The target topology data analysis network is the target model relative to the target teacher model in the model migration, that is, the student model. It improves its own performance by transferring knowledge from the target teacher model. It is a smaller and lightweight model that improves its own performance by transferring knowledge from the teacher model. The student model usually has fewer parameters and computational complexity and is suitable for resource-constrained environments or application scenarios that require efficient reasoning. The target topology data analysis network is used to mine eigenvectors.

[0056] Calculating a first error index based on the first eigenvector and the second eigenvector, the first error index being used to reflect the difference between the first eigenvector and the second eigenvector; wherein the first eigenvector is the output of the target teacher model, that is, the target topological data analysis network performs supervised training based on the output of the target teacher model;

[0057] Based on the first error indicator, the target teacher model and the target topology data analysis network are updated and optimized to form an updated target teacher model and an updated target topology data analysis network. For example, the parameters of the target teacher model and the target topology data analysis network can be updated in the direction of reducing the first error indicator.

[0058] For the above steps, it should be noted that since the optimization targets of the target teacher model and the target topology data analysis network are both the characteristic vectors of the network topology data, the target teacher model and the target topology data analysis network can be jointly optimized to better capture the characteristic vectors of the network topology data; and the target teacher model and the target topology data analysis network are jointly optimized based on the same network topology data, so that the training data domains of the target teacher model and the target topology data analysis network are the same, thereby improving the collaborative effect of the target teacher model and the target topology data analysis network, and then improving the generalization effect of the target teacher model and the target topology data analysis network.

[0059] Specifically, in one embodiment, the training first data includes training a first topology data segment set and topology data segment fusion data. Based on this, the step of determining the training first data for training the first network topology data may include:

[0060] The first network topology data for training is perturbed to form perturbed first network topology data for training the first network topology data; illustratively, some words in the first network topology data for training can be replaced with synonyms or related words, and the order of words or sentences in the first network topology data for training can be adjusted, etc.; wherein, by performing data perturbation on the first network topology data for training, the amount of training data can be increased (it should be noted here that due to the very high data complexity of network topology data, in practical applications, it is difficult to obtain more network topology data for training. Therefore, the amount of training data can be expanded by adding perturbations to a small amount of network topology data. This can orderly reduce the cost of training data, which is particularly important for complex application scenarios such as open-pit mine 5G networks), thereby improving the generalization effect of the target teacher model and the target topology data analysis network;

[0061] Segmenting the perturbed first network topology data to form a set of training first topology data segments of the perturbed first network topology data. As described above, based on the different granularities of the topology data segments, the specific method of data segmentation is also different, such as segmenting by words or sentences (i.e., dividing the perturbed first network topology data);

[0062] Based on the training first topology data fragment set of the perturbed first network topology data, the topology data fragment fusion data of the perturbed first network topology data is determined to form the topology data fragment fusion data of the perturbed first network topology data. The training first topology data fragment set includes the corresponding embedding vectors of the multiple training first topology data fragments (that is, after data segmentation to obtain multiple training first topology data fragments, each training first topology data fragment can be embedded separately to obtain the corresponding embedding vector, and the training first topology data fragment set can be formed by sorting). The topology data fragment fusion data is obtained by fusing the embedding vectors of the multiple training first topology data fragments. For example, the embedding vectors of the multiple training first topology data fragments can be summed or weighted summed to obtain the corresponding topology data fragment fusion data.

[0063] For example:

[0064] The first network topology data for training can be "topology data A represents an open-pit mine 5G network including 5 base stations, each base station has 3 antennas for providing network coverage and communication connection". It is perturbed to obtain the perturbed first network topology data as "topology data A represents an open-pit mine 5G network including 5 base stations, for providing network coverage and communication connection, each base station has 3 antennas". The first network topology data for training is segmented to obtain three training first topology data segments: "topology data A represents an open-pit mine 5G network including 5 base stations", "each base station has 3 antennas", and "used to provide network coverage and communication connection". The first network topology data for perturbation is segmented to obtain three perturbation first topology data segments: "topology data A represents an open-pit mine 5G network including 5 base stations", "used to provide network coverage and communication connection", and "each base station has 3 antennas". Assume that:

[0065] The embedding vector for “topology data A represents a 5G network in an open-pit mine with 5 base stations” is: [0.12, 0.45, -0.78, -0.34, 0.67, 0.91, 0.56, -0.23, 0.78, -0.11, 0.87, -0.65, 0.77, -0.43, 0.09, 0.91, 0.32, -0.54, 0.21, 0.76, -0.89];

[0066] The embedding vector for “each base station has 3 antennas” is: [0.23, -0.45, 0.67, -0.56, 0.78, -0.32, 0.12, 0.87, -0.34, -0.67, 0.91, 0.56, 0.09, 0.77, -0.43, 0.88, -0.11, 0.32, 0, 0, 0];

[0067] The embedding vector corresponding to “used to provide network coverage and communication connectivity” is: [-0.34, 0.67, 0.91, 0.56, -0.23, 0.78, 0.77, -0.43, 0.09, 0.91, 0.32, -0.54, 0.21, 0.76, -0.89, -0.67, 0.56, 0.99, 0.88, 0.12, -0.45].

[0068] Based on the above exemplary embedding vectors, the training first topological data segment set can be: {[0.12, 0.45, -0.78, -0.34, 0.67, 0.91, 0.56, -0.23, 0.78, -0.11, 0.87, -0.65, 0.77, -0.43, 0.09, 0.91, 0.32, -0.54, 0.21, 0.76, -0.89], [0.23, -0.45, 0.67, -0.56, 0.78, -0.32, 0.12, 0.87, -0.34, -0.67, 0.91, 0.56, 0.09, 0.77, -0.43, 0.88, -0.11, 0.32, 0, 0, 0] , [-0.34, 0.67, 0.91, 0.56, -0.23, 0.78, 0.77, -0.43, 0.09, 0.91, 0.32, -0.54, 0.21, 0.76, -0.89, -0.67, 0.56, 0.99, 0.88, 0.12, -0.45]}, the topological data fragment fusion data is (the sum of each embedding vector): [-0.02, 0.67, 0.8, -0.34, 1.22, 1.37, 1.45, 0.21, 1.53, -0.87, 2.1, -0.63, 1.07, 0.1, -1.23, 1.12, 0.77, 0.77, 0.09, 0.88, -1.22].

[0069] In detail, in one embodiment, the first feature vector includes a first topological data vector corresponding to the first topological data segment set for training and a second topological data vector corresponding to the topological data segment fusion data (i.e., the processing result of the target teacher model), and the second feature vector includes a third topological data vector corresponding to the first topological data segment set for training and a fourth topological data vector corresponding to the topological data segment fusion data (i.e., the processing result of the target topological data analysis network). Based on this, the step of calculating the first error index according to the first feature vector and the second feature vector may include:

[0070] Calculating a first local error index based on the first topological data vector and the third topological data vector, wherein the first local error index is used to reflect the difference between the first topological data vector and the third topological data vector, that is, from one perspective, supervised learning of the target topological data analysis network is performed based on the processing results of the target teacher model;

[0071] Calculating a second local error index based on the second topological data vector and the fourth topological data vector, where the second local error index is used to reflect the difference between the second topological data vector and the fourth topological data vector, that is, from another perspective, supervised learning is performed on the target topological data analysis network based on the processing results of the target teacher model;

[0072] Based on the first local error index and the second local error index, a first error index is calculated. For example, the first local error index and the second local error index can be summed or weighted to obtain an error index. In this way, not only the error caused by the characteristic vector of the training first topological data segment set is considered, but also the error caused by the characteristic vector of the topological data segment fusion data is considered. The first error index determined in this way is more comprehensive and more accurate, which can improve the training effect based on the first error index.

[0073] Specifically, in one embodiment, the step of calculating the first local error index based on the first topological data vector and the third topological data vector may include:

[0074] Extracting, based on the set coordinates of at least one covered training first topology data segment in the training first topology data segment set, a local first topology data vector corresponding to the set coordinates from the first topology data vector, and extracting a local third topology data vector corresponding to the set coordinates from the third topology data vector;

[0075] Based on the local first topological data vector and the local third topological data vector, the corresponding first local error index is calculated; since the target topological data analysis network is used to predict the feature vector of the covered training first topological data segment based on the context, and the local first topological data vector and the local third topological data vector respectively represent the feature vectors of the covered training first topological data segment before and after covering, and then the error index is determined based on these two feature vectors, so that the error index can effectively represent the difference between the feature vector predicted by the target topological data analysis network and the feature vector before covering, and then the target topological data analysis network is trained based on the error index, which can improve the accuracy and effectiveness of the training. In addition, the local first topological data vector and the local third topological data vector can be processed based on error calculation functions such as Cross-entropy.

[0076] For example, the local first topology data vector is [1, 0, 0, 0], and the local third topology data vector is [0.8, 0.1, 0.05, 0.05]. Based on this, the error is calculated as:

[0077] L=-Σ(y*log(p))=-(1*log(0.8)+0*log(0.1)+0*log(0.05)+0*log(0.05))

[0078] =-(-0.223+0+0+0)

[0079] =0.223.

[0080] In detail, in one embodiment, the disturbed first network topology data may include disturbed first network topology data having first disturbance information and disturbed first network topology data having second disturbance information (that is, two different disturbances are performed to obtain two different disturbed first network topology data). Based on this, the step of calculating the second local error index according to the second topology data vector and the fourth topology data vector may include:

[0081] Calculating a first disturbance error index based on the first disturbance topology data vector and the second disturbance topology data vector, where the first disturbance topology data vector is a second topology data vector corresponding to the disturbed first network topology data having the first disturbance information, and the second disturbance topology data vector is a fourth topology data vector corresponding to the disturbed first network topology data having the second disturbance information, and the first disturbance error index is used to reflect the difference between the first disturbance topology data vector and the second disturbance topology data vector;

[0082] Calculating a second disturbance error index based on the third disturbance topology data vector and the fourth disturbance topology data vector, where the third disturbance topology data vector is a second topology data vector corresponding to the disturbed first network topology data having the second disturbance information, and the fourth disturbance topology data vector is a fourth topology data vector corresponding to the disturbed first network topology data having the first disturbance information, and the second disturbance error index is used to reflect the difference between the third disturbance topology data vector and the fourth disturbance topology data vector;

[0083] Based on the first disturbance error index and the second disturbance error index, a second local error index is calculated; illustratively, the first disturbance error index and the second disturbance error index can be averaged or weighted averaged to obtain the second local error index; since the network topology data changes in form but does not change in semantics after data perturbation, the feature vectors corresponding to the two disturbed first network topology data are crossed to determine the error index, and then the target topology data analysis network trained based on the error index can better extract the feature vectors of the network topology data, thereby improving the accuracy and effectiveness of network training.

[0084] Specifically, in one embodiment, the step of updating and optimizing the target teacher model and the target topology data analysis network according to the first error indicator to form an updated target teacher model and an updated target topology data analysis network may include:

[0085] updating the network parameters of the target topology data analysis network according to the first error indicator to form an updated target topology data analysis network;

[0086] Based on the network parameters of the updated target topology data analysis network, the target network parameters of the target topology data analysis network are calculated, and the target network parameters are equal to the weighted sum of the network parameters of the updated target topology data analysis network and the target network parameters calculated in the previous round of updating. For example, for the first round of updating, the network parameters of the updated target topology data analysis network can be directly used as the target network parameters of the first round. For the second round of updating, the network parameters of the updated target topology data analysis network and the target network parameters of the first round can be weightedly summed to obtain the target network parameters of the second round. For the third round of updating, the network parameters of the updated target topology data analysis network and the target network parameters of the third round can be weightedly summed to obtain the target network parameters of the third round, and so on. The sum of the weighted coefficients corresponding to the network parameters of the updated target topology data analysis network and the weighted coefficients corresponding to the target network parameters calculated in the previous round of updating is equal to 1, that is, weighted averaging is achieved, making the parameter updating process smoother and more stable.

[0087] According to the target network parameters, the network parameters of the target teacher model are updated to form an updated target teacher model; that is, the network parameters of the target teacher model are replaced with the target network parameters to obtain the updated target teacher model.

[0088] In detail, in one embodiment, the target topology data analysis network includes a shared focusing unit, a first feature mining unit, a second feature mining unit, and a joint feature mining unit. The shared focusing unit is used to perform focused feature mining on the loaded data. The first feature mining unit, the second feature mining unit, and the joint feature mining unit are respectively used to mine the corresponding focused feature mining results to form corresponding feature vectors (as described above). Based on this, according to the first error indicator, the target teacher model and the target topology data analysis network are updated and optimized to form the updated target teacher model and the updated target topology data analysis network. The steps may include:

[0089] Based on the first error index, the shared focusing unit and the first feature mining unit in the target topology data analysis network are updated and optimized. Since the calculation of the first error index is only based on the training of the first network topology data, that is, it is only related to the shared focusing unit and the first feature mining unit, only the shared focusing unit and the first feature mining unit need to be updated and optimized.

[0090] Specifically, in one embodiment, in order to train the second feature mining unit to achieve update optimization, the data simulation method based on the open-pit mine 5G network may further include:

[0091] The training second network topology data is segmented using the topology data segmentation model to form corresponding third feature vectors, where the third feature vectors include feature vectors of respective multiple training second topology data segments (such as the aforementioned data segments) formed by segmenting the training second network topology data; illustratively, the topology data segmentation model may be a target teacher model or may not be a target teacher model, but may have the same network architecture;

[0092] Overwriting at least one training second topology data segment in the training second topology data segment set corresponding to the third eigenvector to form a corresponding fourth eigenvector, as described above, that is, overwriting at least one training second topology data segment in the training second topology data segment set, and then processing the overwritten data to obtain the fourth eigenvector;

[0093] Loading the fourth eigenvector into the target topological data analysis network and mining the corresponding fifth eigenvector, i.e., further mining through the target topological data analysis network;

[0094] Extracting the local second topology data vector corresponding to the set coordinates of the at least one covered training second topology data segment in the training second topology data segment set from the fifth eigenvector, as described above;

[0095] Calculating a second error index based on the local second topological data vector and the local fourth topological data vector, where the local fourth topological data vector is a feature vector of at least one training second topological data segment included in the fourth feature vector, and the second error index is used to reflect the difference between the local second topological data vector and the local fourth topological data vector, as described above.

[0096] Based on the second error indicator, the second feature mining unit in the target topology data analysis network is updated and optimized, as described above. Since the second error indicator is obtained based on the feature vector of the training second network topology data, and the feature vector of the training second network topology data is obtained based on the second feature mining unit in the target topology data analysis network, the second feature mining unit is trained based on the second error indicator, thereby improving the effectiveness and accuracy of the training. Moreover, after the training second network topology data is covered, the target topology data analysis network is trained, so that the target topology data analysis network can learn the contextual information in the training second network topology data, and thus can perform semantic estimation on the training second network topology data.

[0097] It is understandable that the target topology data analysis network may not include the second feature mining unit, that is, both network topology data are processed by a single feature mining unit. Alternatively, when two feature mining units are included, the two feature mining units can be configured such that different feature mining units process network topology data with different tendencies. For example, the network topology data learned by the first feature mining unit is of low engineering effort, while the network topology data learned by the second feature mining unit is of low equipment cost.

[0098] In detail, in one embodiment, since in the aforementioned example, the first network topology data and the second network topology data are trained separately, and the target topology data analysis network ultimately needs to aggregate the feature vectors of the two network topology data, based on this, the data simulation method based on the open-pit mine 5G network may further include:

[0099] Determine a training network topology data combination, where the training network topology data combination includes two network topology data corresponding to simulation results that meet preset simulation requirements;

[0100] Using the target teacher model, a sixth eigenvector is determined, and using the topological data segmentation model, a seventh eigenvector is determined, the sixth eigenvector being the eigenvector of the first training network topology data in the training network topology data combination, and the seventh eigenvector being the eigenvector of the second training network topology data in the training network topology data combination, as described above;

[0101] Overlay at least one training first topology data segment in the training first topology data segment set corresponding to the sixth eigenvector, and mine the corresponding eighth eigenvector, as above;

[0102] Overlay at least one training second topology data segment in the training second topology data segment set corresponding to the seventh eigenvector, and mine the corresponding ninth eigenvector, as above;

[0103] The eighth eigenvector and the ninth eigenvector are loaded into the target topology data analysis network respectively to form a training aggregated eigenvector corresponding to the training network topology data combination;

[0104] Calculate a third error index based on the training aggregated feature vector, the sixth feature vector, and the seventh feature vector. The third error index is used to reflect the difference between the feature vector of the first training network topology data and the sixth feature vector in the training aggregated feature vector (calculated as before), and the difference between the feature vector of the second training network topology data and the seventh feature vector in the training aggregated feature vector (calculated as before).

[0105] According to the third error index, the target topology data analysis network is updated and optimized; that is, the network parameters of the target topology data analysis network can be updated in the direction of reducing the third error index, that is, the network parameters of the shared focus unit, the first feature mining unit, the second feature mining unit and the joint feature mining unit included in the target topology data analysis network need to be updated and optimized, that is, the first feature mining unit and the second feature mining unit can be updated separately first, and then the shared focus unit, the first feature mining unit, the second feature mining unit and the joint feature mining unit can be updated as a whole. In this way, small-grained updates can be performed to ensure the independence and accuracy of each unit, and the coordination of each unit can be guaranteed.

[0106] In detail, in one embodiment, the target topology data analysis network includes multiple network hierarchical structures, the first network hierarchical structure in the multiple network hierarchical structures includes a shared focusing unit, a second feature mining unit, and a first feature mining unit, and the second network hierarchical structure in the multiple network hierarchical structures includes a shared focusing unit, a second feature mining unit, a first feature mining unit, and a joint feature mining unit. Based on this, the eighth feature vector and the ninth feature vector are loaded into the target topology data analysis network respectively to form a training aggregate feature vector corresponding to the training network topology data combination, which may include:

[0107] The eighth eigenvector is sequentially loaded into the shared focusing unit and the first feature mining unit in the first network hierarchy, and the corresponding first candidate topology vector for training is output; that is, the eighth eigenvector can be first processed by the shared focusing unit, and then the result of the processing can be further processed by the first feature mining unit to obtain the first candidate topology vector for training;

[0108] The ninth eigenvector is sequentially loaded into the shared focusing unit and the second feature mining unit in the first network hierarchy structure, and the corresponding second candidate topology vector for training is output; that is, the ninth eigenvector can be first processed by the shared focusing unit, and then the result of the processing can be further processed by the second feature mining unit to obtain the second candidate topology vector for training;

[0109] The first candidate topology vector for training is loaded into the shared focus unit and the joint feature mining unit in the second network hierarchy structure in sequence, and the second candidate topology vector for training is loaded into the shared focus unit and the joint feature mining unit in the second network hierarchy structure in sequence. The first candidate topology vector for training and the second candidate topology vector for training are aggregated by using the joint feature mining unit to form a training aggregate feature vector corresponding to the training network topology data combination. Exemplarily, the first candidate topology vector for training is loaded into the shared focus unit and the joint feature mining unit in the second network hierarchy structure in sequence for processing to obtain a first depth vector, the second candidate topology vector for training is loaded into the shared focus unit and the joint feature mining unit in the second network hierarchy structure in sequence for processing to obtain a second depth vector, and the first depth vector and the second depth vector are aggregated, such as splicing (cascading), to form a training aggregate feature vector corresponding to the training network topology data combination.

[0110] It can be understood that, as mentioned above, the target topology data analysis network can include an encoding subnetwork (Encoder) and a decoding subnetwork (Decoder). The above-mentioned network update process is aimed at the encoding subnetwork. For the update of the decoding subnetwork, on the basis of the training network topology data combination, the training network topology data combination can also be configured with corresponding labels (which is also a type of network topology data, which is determined based on the two training network topology data included in the training network topology data combination, such as one of the two training network topology data has a smaller engineering volume and the other has a smaller equipment cost, and the network topology data as a label can integrate the engineering volume and equipment cost, such as the engineering volume and equipment cost are between the engineering volume and equipment cost of the two training network topology data). In this way, after obtaining the training aggregated feature vector based on the encoding subnetwork, it can be decoded (generated) based on the decoding subnetwork to obtain the predicted network topology data. Then, the error between the predicted network topology data and the label is calculated, so that the network parameters of the decoding subnetwork can be updated and optimized based on the error.

[0111] Combine Figure 3 The embodiment of the present invention further provides a data simulation device based on the open-pit mine 5G network, which can be applied to the above-mentioned data simulation system based on the open-pit mine 5G network. The data simulation device based on the open-pit mine 5G network may include:

[0112] A network topology data acquisition module 10 is configured to obtain two network topology data sets, each of which is used to characterize the network layout of the open-pit mine 5G network and the simulation results of which are used to characterize the performance indicators of the open-pit mine 5G network.

[0113] a data segment set determining module 20, configured to construct a first topology data segment set for a first network topology data of the two network topology data and a second topology data segment set for a second network topology data of the two network topology data, wherein each first topology data segment in the first topology data segment set belongs to the first network topology data, and each second topology data segment in the second topology data segment set belongs to the second network topology data;

[0114] a feature vector determination module 30 for analyzing a network using target topology data, and determining a feature vector of the first network topology data and a feature vector of the second network topology data based on a first set of topology data segments of the first network topology data and a second set of topology data segments of the second network topology data;

[0115] The network topology data determination module 40 is used to analyze the network using the target topology data, generate new network topology data based on the aggregation result of the feature vector of the first network topology data and the feature vector of the second network topology data, and determine the end of the simulation when the simulation result of the new network topology data meets the preset simulation requirements, and use the new network topology data as the target network topology data.

[0116] In summary, the data simulation method and system based on the open-pit mine 5G network provided by the present invention can obtain two network topology data; construct a first topology data fragment set and a second topology data fragment set of the two network topology data; based on the first topology data fragment set and the second topology data fragment set, determine the characteristic vector of the first network topology data and the characteristic vector of the second network topology data; based on the aggregation result of the characteristic vector of the first network topology data and the characteristic vector of the second network topology data, generate new network topology data, and when the simulation result of the new network topology data meets the preset simulation requirements, determine that the simulation is over, and use the new network topology data as the target network topology data. Based on the foregoing content, after obtaining the two network topology data whose simulation results of the network simulation for the open-pit mine 5G network meet the preset simulation requirements, it is not simply to determine one network topology data from the two network topology data, but to mine the two topology data, and then fuse the mined feature vectors, so that new network topology data can be generated based on the fusion results. In this way, the new network topology data can take into account the two obtained network topology data. Therefore, when the simulation results of the new network topology data meet the preset simulation requirements, the reliability of the new network topology data as the target network topology data can be guaranteed, so that the reliability of the network topology data determined during the data simulation process can be improved, thereby improving the problem of relatively low reliability in the existing technology.

[0117] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A data simulation method based on 5G network in open pit mines, characterized in that: include: Acquire two network topology data whose simulation results for the open-pit mine 5G network meet preset simulation requirements. The network topology data is used to characterize the network layout in the open-pit mine 5G network, and the simulation results are used to characterize the performance indicators of the open-pit mine 5G network. Constructing a first topology data segment set for a first network topology data of the two network topology data and a second topology data segment set for a second network topology data of the two network topology data, wherein each first topology data segment in the first topology data segment set belongs to the first network topology data, and each second topology data segment in the second topology data segment set belongs to the second network topology data; Analyzing the network using the target topology data, and determining a feature vector of the first network topology data and a feature vector of the second network topology data based on a first topology data segment set of the first network topology data and a second topology data segment set of the second network topology data; The target topology data is used to analyze the network, and new network topology data is generated based on the aggregation result of the feature vector of the first network topology data and the feature vector of the second network topology data. When the simulation result of the new network topology data meets the preset simulation requirements, the simulation is determined to be ended, and the new network topology data is used as the target network topology data.

2. The data simulation method based on the open-pit mine 5G network according to claim 1, characterized in that: The step of analyzing the network using the target topology data and determining the feature vector of the first network topology data and the feature vector of the second network topology data based on the first topology data segment set of the first network topology data and the second topology data segment set of the second network topology data includes: Loading the first topology data segment set into a target topology data analysis network, and mining a corresponding first candidate topology vector using a shared focusing unit and a first feature mining unit in the target topology data analysis network; Loading the second topology data segment set into the target topology data analysis network, and mining the corresponding second candidate topology vector using the shared focusing unit and the second feature mining unit in the target topology data analysis network; The first candidate topology vector and the second candidate topology vector are loaded into a shared focusing unit and a joint feature mining unit included in the target topology data analysis network, and the focused feature mining results of the first candidate topology vector and the focused feature mining results of the second candidate topology vector are aggregated by the joint feature mining unit to form a corresponding aggregated feature vector, wherein the aggregated feature vector includes the feature vector of the first network topology data and the feature vector of the second network topology data; The shared focusing unit is used to perform focused feature mining on the loaded data, and the first feature mining unit, the second feature mining unit and the joint feature mining unit are respectively used to mine the corresponding focused feature mining results to form corresponding feature vectors.

3. The data simulation method based on the open-pit mine 5G network according to claim 1 or 2, characterized in that: The data simulation method based on the open-pit mine 5G network also includes: Determine training first data for training first network topology data, where the training first data is used to indicate a training first topology data segment set, and the training first topology data segment set includes a plurality of training first topology data segments in the training first network topology data; Loading the first training data into the target teacher model and outputting the corresponding first eigenvector; Overwriting at least one training first topology data segment in the training first topology data segment set indicated by the training first data to form corresponding training second data; Loading the second training data into a target topological data analysis network and outputting a corresponding second eigenvector, wherein the target topological data analysis network is a target model relative to the target teacher model in the model migration, and improves its own performance by performing knowledge migration from the target teacher model, and the target topological data analysis network is used to mine the eigenvector; Calculating a first error index based on the first eigenvector and the second eigenvector, where the first error index is used to reflect the difference between the first eigenvector and the second eigenvector; According to the first error indicator, the target teacher model and the target topology data analysis network are updated and optimized to form an updated target teacher model and an updated target topology data analysis network.

4. The data simulation method based on the open-pit mine 5G network according to claim 3 is characterized in that: The first training data includes a first training topology data segment set and topology data segment fusion data. The step of determining the first training data for training the first network topology data includes: Performing data perturbation on the training first network topology data to form perturbed first network topology data of the training first network topology data; Segmenting the disturbed first network topology data to form a training first topology data segment set of the disturbed first network topology data; Based on the training first topology data segment set of the perturbed first network topology data, the topology data segment fusion data of the perturbed first network topology data is determined to form the topology data segment fusion data of the perturbed first network topology data. The training first topology data segment set includes the corresponding embedding vectors of multiple training first topology data segments, and the topology data segment fusion data is obtained by fusing the embedding vectors of the multiple training first topology data segments.

5. The data simulation method based on the open-pit mine 5G network according to claim 4 is characterized in that: The first feature vector includes a first topological data vector corresponding to the first training topological data segment set and a second topological data vector corresponding to the topological data segment fusion data, the second feature vector includes a third topological data vector corresponding to the first training topological data segment set and a fourth topological data vector corresponding to the topological data segment fusion data, and the step of calculating a first error index based on the first feature vector and the second feature vector includes: Calculating a first local error index based on the first topology data vector and the third topology data vector, where the first local error index is used to reflect the difference between the first topology data vector and the third topology data vector; calculating a second local error index based on the second topology data vector and the fourth topology data vector, wherein the second local error index is used to reflect the difference between the second topology data vector and the fourth topology data vector; A first error index is calculated according to the first local error index and the second local error index.

6. The data simulation method based on the open-pit mine 5G network according to claim 5, characterized in that: The step of calculating a first local error index based on the first topological data vector and the third topological data vector includes: Extracting, based on the set coordinates of the at least one covered training first topology data segment in the training first topology data segment set, a local first topology data vector corresponding to the set coordinates in the first topology data vector, and extracting a local third topology data vector corresponding to the set coordinates in the third topology data vector; A corresponding first local error index is calculated based on the local first topology data vector and the local third topology data vector.

7. The data simulation method based on the open-pit mine 5G network according to claim 3 is characterized in that The target topology data analysis network includes a shared focusing unit, a first feature mining unit, a second feature mining unit, and a joint feature mining unit. The shared focusing unit is used to perform focused feature mining on the loaded data. The first feature mining unit, the second feature mining unit, and the joint feature mining unit are respectively used to mine the corresponding focused feature mining results to form corresponding feature vectors. The data simulation method based on the open-pit mine 5G network also includes: Using a topology data segmentation model, segmenting the training second network topology data to form corresponding third feature vectors, wherein the third feature vectors include feature vectors of respective multiple training second topology data segments formed by segmenting the training second network topology data; Overlaying at least one training second topology data segment in the training second topology data segment set corresponding to the third eigenvector to form a corresponding fourth eigenvector; Loading the fourth eigenvector into the target topology data analysis network to mine the corresponding fifth eigenvector; Extracting, according to the set coordinates of the at least one covered training second topology data segment in the training second topology data segment set, a local second topology data vector corresponding to the set coordinates in the fifth eigenvector; Calculating a second error index based on the local second topological data vector and the local fourth topological data vector, wherein the local fourth topological data vector is a feature vector of the at least one training second topological data segment included in the fourth feature vector, and the second error index is used to reflect the difference between the local second topological data vector and the local fourth topological data vector; According to the second error indicator, the second feature mining unit in the target topology data analysis network is updated and optimized.

8. The data simulation method based on the open-pit mine 5G network according to claim 3 is characterized in that: The target topology data analysis network includes a shared focusing unit, a first feature mining unit, a second feature mining unit, and a joint feature mining unit. The shared focusing unit is used to perform focused feature mining on the loaded data. The first feature mining unit, the second feature mining unit, and the joint feature mining unit are respectively used to mine the corresponding focused feature mining results to form corresponding feature vectors. The data simulation method based on the open-pit mine 5G network also includes: Determining a training network topology data combination, wherein the training network topology data combination includes two network topology data corresponding to simulation results that meet preset simulation requirements; Using the target teacher model, a sixth eigenvector is determined, and using the topological data segmentation model, a seventh eigenvector is determined, wherein the sixth eigenvector is a eigenvector of the first training network topology data in the training network topology data combination, and the seventh eigenvector is a eigenvector of the second training network topology data in the training network topology data combination; Overlaying at least one training first topology data segment in the training first topology data segment set corresponding to the sixth eigenvector, and mining the corresponding eighth eigenvector; Overlaying at least one training second topology data segment in the training second topology data segment set corresponding to the seventh eigenvector, and mining the corresponding ninth eigenvector; loading the eighth eigenvector and the ninth eigenvector into the target topology data analysis network respectively to form a training aggregated eigenvector corresponding to the training network topology data combination; Calculating a third error index based on the training aggregated feature vector, the sixth feature vector, and the seventh feature vector, wherein the third error index is used to reflect the difference between the feature vector of the first training network topology data in the training aggregated feature vector and the sixth feature vector, and the difference between the feature vector of the second training network topology data in the training aggregated feature vector and the seventh feature vector; The target topology data analysis network is updated and optimized according to the third error indicator.

9. The data simulation method based on the open-pit mine 5G network according to claim 8, characterized in that: The target topology data analysis network includes multiple network hierarchical structures, a first network hierarchical structure among the multiple network hierarchical structures includes a shared focusing unit, a second feature mining unit, and a first feature mining unit, and a second network hierarchical structure among the multiple network hierarchical structures includes a shared focusing unit, a second feature mining unit, a first feature mining unit, and a joint feature mining unit, and the steps of respectively loading the eighth feature vector and the ninth feature vector into the target topology data analysis network to form a training aggregated feature vector corresponding to the training network topology data combination include: sequentially loading the eighth feature vector into the shared focus unit and the first feature mining unit in the first network hierarchical structure, and outputting the corresponding first candidate topology vector for training; sequentially loading the ninth feature vector into the shared focusing unit and the second feature mining unit in the first network hierarchical structure, and outputting the corresponding second candidate topology vector for training; The training first candidate topology vector is sequentially loaded into the shared focus unit and the joint feature mining unit in the second network hierarchical structure, and the training second candidate topology vector is sequentially loaded into the shared focus unit and the joint feature mining unit in the second network hierarchical structure. The training first candidate topology vector and the training second candidate topology vector are aggregated by using the joint feature mining unit to form a training aggregated feature vector corresponding to the training network topology data combination.

10. A data simulation system based on 5G network in open pit mines, characterized in that: It includes a processor and a memory, the memory is used to store a computer program, and the processor is used to execute the computer program to implement the data simulation method based on the open-pit mine 5G network as described in any one of claims 1 to 9.

Images