| CPC G01G 19/03 (2013.01) [G01M 5/0008 (2013.01)] | 10 Claims |
|
1. A regional traffic heavy load digital weighing method, comprising:
Step S100: based on data of a physical weighing system in a region and basic data of a bridge in the region, selecting a number M of reference physical weighing systems MVW[i, 0], 1≤i≤M, and determining a number N of bridge group transfer layers;
Step S200: based on each of the number M of reference physical weighing systems, constructing a corresponding single digital bridge scale model MVW[i, 1];
Step S300: obtaining branch scale network models MVW{[i, j]} based on the reference physical weighing systems through migration learning and correction, 1≤i≤M, j=1, 2, . . . N;
Step S400: composing a regional digital scale network model MVW{[i, j]} based on branch scale network models corresponding to different reference physical weighing systems, i=1, 2, . . . , M, j=1, 2, . . . , N;
wherein the Step S100 comprises:
Step S110: based on the provided data, drawing an overall traffic bridge network distribution in the region;
Step S120: marking bridge groups with high technical condition ratings and corresponding weighing systems in a vicinity;
Step S130: based on a principle of minimizing error transmission and a principle of minimizing cost of model consumption, selecting weighing systems from the weighing systems marked in the step S120 as the reference physical weighing systems MVW[i, 0];
wherein the step S200 comprises:
Step S210: constructing a structural response monitoring system at a key position of the bridge and a positive lateral traffic splicing identification system before the vehicle is on the bridge, to obtain a spatio-temporal matrix of the deformation data Deformation(i, j) and a vehicle feature input information matrix Info_in(i, j), respectively;
Step S220: obtaining a vehicle feature output information matrix Info_out(i, j) based on a data fusion of the step S210;
Step S230: passing, through a license plate number, the vehicle feature output information matrix Info_out(i, 0) corresponding to the reference physical weighing systems MVW [i, 0] to the positive lateral traffic splicing identification system as an output of the single digital bridge scale model, and using response data of the structural response monitoring system as an input of the single digital bridge scale model, to obtain a dataset [MVW(i, 0), Deformation(i, 1)] of the reference physical weighing systems corresponding to an entity bridge;
Step S240: using a deep learning model to train the dataset [MVW(i, 0), Deformation(i, 1)] and inversing to obtain the single digital bridge scale model MVW[i, 1];
wherein MVW(i, 0) denotes a weighing result of the reference physical weighing system;
wherein the Step S300 comprises:
Step S310: obtaining a reference model of the single digital bridge scale model MVW[i, j−1] by replicating the single digital bridge scale model MVW[i, j] through migration learning; wherein, 2≤j≤N;
S320: using the dataset [MVW(i, j−1), Deformation(i, j)]transferred by the license plate number to correct the reference model to obtain the corrected single digital bridge scale model MVW[i, j] and a corresponding weighing result MVW(i, j);
S330: by analogy, forming the branch scale network model MVW{[i, j]} driven by the reference physical weighing system MVW[i, 0];
wherein in the Step S400, when there a are cross sections between different branch scale network models in the regional digital scale network model, branch scale network models with cross sections need to be verified.
|