Production of a turbomachine vane
A method for designing a turbomachine vane, in which predefined input parameters are transmitted to a neuronal network system and vane parameters are determined and output by the neuronal network system based on the transmitted input parameters. The neuronal network system has several separate neuronal networks each with an output layer, each of which determines one or more of the vane parameters and outputs same via the output layer. A first neuronal network and a second neuronal network belong to the separate neuronal networks of the neuronal network system and the vane parameter(s) which are determined by the first neuronal network and output via the output layer of said neuronal network differ(s) from the vane parameter(s) that are determined by the second neuronal network and are output via the output layer of said neuronal network.
1 . A computer-implemented method for designing a turbomachine blade, comprising:
communicating predefined input parameters to a neural network system;
determining blade parameters by the neural network system based on the communicated predefined input parameters;
outputting said blade parameters;
wherein the neural network system comprises a plurality of separate neural networks including a first neural network to determine at least one of the blade parameters, and a second neural network to determine at least a different one of the blade parameters;
wherein the at least one of the blade parameters is fewer than all of the blade parameters, and the at least the different one of the blade parameters is fewer than all of the blade parameters;
wherein the first neural network and the second neural network are configured to determine fewer blade parameters as compared to the blade parameters being determined by a same neural network;
wherein before the blade parameters are determined, the first neural network and the second neural network are trained with training data, with the training data being a subset of a basic training data set that includes data obtained during earlier designs of turbomachine blades and/or during earlier optimizations of turbomachine blade designs;
determining whether the basic training data set contains redundant parameters from a statistical standpoint and from a physical standpoint, with the statistical standpoint using Pearson correlation coefficient to measure linear correlation between two parameters in the basic training data set,
filtering the redundant parameters so that a number of predefined input parameters required for determining the blade parameters is reduced and a topology of the neural network system is reduced;
wherein the blade parameters determined by the neural network system comprise geometric blade parameters;
wherein the geometric blade parameters comprise at least one curvature parameter and at least one thickness parameter, wherein the first neural network system determines the at least one curvature parameter and the second neural network system determines the as least one thickness parameter, or wherein the geometric blade parameters parameterize a freeform curve or a Bézier curve, which characterizes one or more blade surfaces; and
wherein the predefined input parameters communicated to the neural network system are two-dimensional parameters including a suction-side Mach number distribution a pressure-side Mach number distribution, and at least one pressure side distribution.
2 . The method as claimed in claim 1 ,
wherein the neural network system comprises a plurality of neural connections, and wherein after the neural network system has been trained by the training data and before the blade parameters are determined, at least one of the neural connections is removed by a pruning method.
3 . The method as claimed in claim 1 ,
wherein the blade parameters determined by the neural network system comprise a blade efficiency of the turbomachine blade and/or a logarithmic decrement of the turbomachine blade.
4 . The method as claimed claim 1 ,
wherein the neural network system comprises a third neural network to determine a logarithmic decrement of the turbomachine blade, and a fourth neural network to determine a blade efficiency of the turbomachine blade.
5 . The method as claimed in claim 1 ,
wherein at least one of the first neural network and the second neural network comprises a plurality of intermediate layers.
6 . The method as claimed in claim 1 ,
wherein at least one of the first neural network and the second neural network comprises an ensemble of a plurality of neural subnetworks,
wherein each neural subnetwork determines a blade parameter value for one of the blade parameters, and
wherein the blade parameter values as determined by the ensemble of the plurality of neural subnetworks are averaged in a weighted or an unweighted fashion, wherein each averaged blade parameter value obtained in this way is output as one of the blade parameters.
7 . A method for producing a turbomachine blade using the method of claim 1 , comprising:
optionally modifying the blade parameters based on at least one optimization criterion; and
producing the turbomachine blade in accordance with the blade parameters determined and optionally modified.
8 . A computer for carrying out the method of claim 1 the neural network system configured to receive the predefined input parameters and to determine the blade parameters based on the received predefined input parameters and to output said blade parameters, and
wherein the neural network system comprises the plurality of separate neural networks including the first neural network and the second neural network.
9 . A non-transitory machine-readable storage medium, comprising:
a program code which causes a computer to carry out the method of claim 1 when the program code is executed by the computer.