Flow field prediction device, learning device, flow field prediction program, and learning program
A flow field prediction device includes: an analysis result acquisition unit that acquires an analysis result based on a first computation method of analyzing at least an influence between buildings with respect to a flow field in which a city block including a plurality of buildings is set as a target region; and a prediction processing unit that inputs a new analysis result acquired by the analysis result acquisition unit to a learning model trained using teaching data that includes an analysis result based on a second computation method that is more accurate than the first computation method, as correct answer data for input data including at least the analysis result based on the first computation method, and outputs an output from the learning model as a prediction result of the flow field.
1 . A flow field prediction device comprising:
a memory; and
a processor, wherein the processor is configured to:
provide a machine learning model, by:
receiving a plurality of instances of input data, each instance of input data relating to buildings in a city block including a plurality of buildings;
analyzing each instance of input data according to a low-accuracy computation method, obtaining a low-accuracy analysis result, and setting the low-accuracy analysis result as a first element in paired training data;
analyzing each instance of input data according to a high-accuracy computation method, obtaining a high-accuracy analysis result, and setting the high-accuracy analysis result as a second element in the paired training data; and
training the machine learning model using the paired training data, wherein each low-accuracy analysis result is used as initial data and the high-accuracy analysis result is used as correct answer data;
receive a new input of a new city block including a new plurality of buildings that is set as a target region;
perform the low-accuracy computation method on the new input in order to provide a new low-accuracy analysis result; and
use the machine learning model to apply an adjustment to the new low-accuracy analysis result in order to provide a final prediction, wherein providing the final prediction is performed without performing the high-accuracy computation method on the new input.
2 . The flow field prediction device according to claim 1 , wherein the processor is further configured to:
acquire, as the new low-accuracy analysis result, a plurality of pieces of plane data including plane data made up of information on a flow field at a predetermined height in the city block, and
with the machine learning model, output at least a prediction result of the predetermined height in response to the input of the new low-accuracy analysis result.
3 . The flow field prediction device according to claim 2 , wherein the processor is further configured to:
acquire, as the new low-accuracy analysis result, the plurality of pieces of plane data of successive heights within a predetermined range, and
with the machine learning model, output a plurality of prediction results of successive heights in the same range as the input in response to the input of the new low-accuracy analysis result.
4 . The flow field prediction device according to claim 2 , wherein the processor is further configured to:
acquire, as the new low-accuracy analysis result, the plurality of pieces of plane data of successive heights within a predetermined range, and
with the machine learning model, sequentially add, to the input, outputs successive in a height direction obtained by input of the plane data, and thereby output a plurality of prediction results of successive heights.
5 . The flow field prediction device according to claim 2 , wherein the processor is further configured to:
acquire, as the new low-accuracy analysis result, a plurality of pieces of plane data of successive points in time, and
with the learning model, sequentially add, to the input, outputs successive in a time direction obtained by input of the plane data, and thereby output a plurality of prediction results of successive points in time.
6 . The flow field prediction device according to claim 1 , wherein
the new low-accuracy analysis result based on the first computation method includes information on at least one of wind separation, wake, or contraction.
7 . The flow field prediction device according to claim 1 , wherein
the low-accuracy computation method is Reynolds-Averaged Navier-Stokes Simulation (RANS), and
the high-accuracy computation method is Large Eddy Simulation (LES).
8 . The flow field prediction device according to claim 1 , wherein the machine learning model is configured by a Convolutional Neural Network (CNN).
9 . A non-transitory computer-readable medium storing a flow field prediction program for causing a computer to function as the flow field prediction device according to claim 1 .