Systems and methods for building dynamic reduced order physical models
Methods and apparatuses that generate a simulation object for a physical system are described. The simulation object includes a trained computing structure to determine future output data of the physical system in real time. The computing structure is trained with a plurality of input units and one or more output units. The plurality of input units include regular input units to receive input data and output data of the physical system. The output units include one or more regular output units to predict a dynamic rate of change of the input data over a period of time. The input data and output data of the physical system are obtained for training the computing structure. The input data represent a dynamic input excitation to the physical system over the period of time. And the output data represents a dynamic output response of the physical system to the dynamic input excitation over the period of time.
1. A computer-implemented method comprising:
obtaining input data and output data of a physical system, the input data representing a dynamic input excitation to the physical system over a period of time, and the output data representing a dynamic output response of the physical system to the dynamic input excitation over the period of time;
configuring a neural network including an input layer and an output layer, the input layer including a plurality of input units to receive the input data and the output data, and the output layer including one or more output units;
training the neural network with the input data and the output data of the physical system to predict a dynamic rate of change of the input data via the one or more output units over the period of time, the input layer updated with a free input unit in additional to the plurality of input units, the output layer updated with a free output unit in addition to the one or more output units, the free output unit to predict a rate of change of data received by the free input unit over the period of time; and
generating a simulation object for the physical system, the simulation object including the trained neural network to determine future output data of the physical system in real time.
2. The computer-implemented method of claim 1 , wherein the free input unit is not configured to receive the input data or the output data of the plurality of input units and the one or more of output units, and the free input unit is configured to provide a time varied data and a dynamic rate of change of the time varied data during the period of time.
3. The computer-implemented method of claim 2 , further comprising:
generating a predicted dynamic rate of change of the input data over the period of time based on the neural network, the input data, and the output data;
comparing the dynamic rate of change of the input data and the predicted dynamic rate of change of the input data over the period of time; and
generating a prediction residual indicating an accuracy of the predicted dynamic rate of change of the input data based on an accuracy threshold.
4. The computer-implemented method of claim 1 , wherein the free output unit is configured to provide a free output value, the free input unit is configured to receive a free input value, and the training comprises:
assigning the free input value such that the free output value corresponds to a dynamic rate of change of the free input value.
5. The computer-implemented method of claim 4 , wherein the free input value includes an initial value for the period of time, and wherein the training further comprises:
generating the predicted dynamic rate of change of the input data based a selection of the initial value from a range of possible initial values.
6. The computer-implemented method of claim 5 , wherein the initial value is selected based on a back propagation algorithm.
7. The computer-implemented method of claim 1 , wherein the output data is generated based on the input data over the period of time using a model of the physical system.
8. The computer-implemented method of claim 1 , further comprising:
executing a physical solver that models physical properties of the physical system to generate the output data from the input data over the period of time.
9. The computer-implemented method of claim 1 , wherein the neural network comprises a previously configured neural network, the previously configured neural network configured without the free input unit and without the free output unit.
10. The computer-implemented method of claim 9 , wherein the previously configured neural network comprises the plurality of input units and the one or more output units, the method further comprising:
training the previously configured neural network for the one or more output units to predict the dynamic rate of change of the input data according to the input data and the output data received via the plurality of input units over the period of time.
11. A non-transitory computer-readable storage medium storing instructions, which when executed by a computer processing system, cause the computer processing system to perform operations comprising:
obtaining input data and output data of a physical system, the input data representing a dynamic input excitation to the physical system over a period of time, and the output data representing a dynamic output response of the physical system to the dynamic input excitation over the period of time;
configuring a neural network including an input layer and an output layer, the input layer including a plurality of input units to receive the input data and the output data, and the output layer including one or more output units;
training the neural network with the input data and the output data of the physical system to predict a dynamic rate of change of the input data via the one or more output units over the period of time, the input layer updated with a free input unit in additional to the plurality of input units, the output layer updated with a free output unit in addition to the one or more output units, the free output unit to predict a rate of change of data received by the free input unit over the period of time; and
generating a simulation object for the physical system, the simulation object including the trained neural network to determine future output data of the physical system in real time.
12. The non-transitory computer-readable storage medium of claim 11 , wherein the free input unit is not configured to receive the input data or the output data of the plurality of input units and the one or more output units, and the free input unit is configured to provide a time varied data and a dynamic rate of change of the time varied data during the period of time.
13. The non-transitory computer-readable storage medium of claim 12 , the operations further comprising:
generating a predicted dynamic rate of change of the input data over the period of time based on the neural network, the input data, and the output data;
comparing the dynamic rate of change of the input data and the predicted dynamic rate of change of the input data over the period of time; and
generating a prediction residual indicating an accuracy of the predicted dynamic rate of change of the input data based on an accuracy threshold.
14. The non-transitory computer-readable storage medium of claim 11 , wherein the free output unit is configured to provide a free output value, the free input unit is configured to receive a free input value, and the operations for training further comprise:
assigning the free input value such that the free output value corresponds to a dynamic rate of change of the free input value.
15. The non-transitory computer-readable storage medium of claim 14 , wherein the free input value includes an initial value for the period of time, and wherein the training further comprises:
generating the predicted dynamic rate of change of the input data based a selection of the initial value from a range of possible initial values.
16. The non-transitory computer-readable storage medium of claim 11 , wherein the output data is generated based on the input data over the period of time using a model of the physical system.
17. The non-transitory computer-readable storage medium of claim 11 , the operations further comprising:
executing a physical solver that models physical properties of the physical system to generate the output data from the input data over the period of time.
18. A system, comprising:
a memory storing instructions;
one or more processors coupled to the memory, the one or more processors executing the instructions from the memory, the one or more processors configured to perform operations comprising:
obtaining input data and output data of a physical system, the input data representing a dynamic input excitation to the physical system over a period of time, and the output data representing a dynamic output response of the physical system to the dynamic input excitation over the period of time,
configuring a neural network including an input layer and an output layer, the input layer including a plurality of input units to receive the input data and the output data, and the output layer including one or more output units,
training the neural network with the input data and the output data of the physical system to predict a dynamic rate of change of the input data via the one or more output units over the period of time, the input layer updated with a free input unit in additional to the plurality of input units, the output layer updated with a free output unit in addition to the one or more output units, the free output unit to predict a rate of change of data received by the free input unit over the period of time, and
generating a simulation object for the physical system, the simulation object including the trained neural network to determine future output data of the physical system in real time.
19. The system of claim 18 , wherein the free input unit is configured to receive a free input value, the free input value includes an initial value for the period of time, and the operations further comprise:
generating the predicted dynamic rate of change of the input data based a selection of the initial value from a range of possible initial values.
20. The system of claim 18 , wherein the output data is generated based on the input data over the period of time using a model of the physical system.