Enveloped mixture of experts model
A system for monitoring operation of a gas turbine engine comprises a plurality of expert networks each configured to generate a separate gas path parameter responsive to a separate actuator position. A router network configured to generate a weighting vector responsive to at least one ambient condition parameter. The weighting vector includes a plurality of weighting values each associated with one of the plurality of expert networks. Summing circuitry configured to apply the plurality of weighting values to each of the associated separate gas path parameters from the plurality of expert networks and sum each of the plurality of weighted separate gas path parameters from the plurality of expert networks to a weighted sum value.
1 . A system for monitoring operation of a gas turbine engine, comprising:
a plurality of expert networks each configured to generate a separate gas path parameter responsive to a separate actuator position;
a router network configured to generate a weighting vector responsive to at least one ambient condition parameter, wherein the weighting vector includes a plurality of weighting values each associated with one of the plurality of expert networks; and
summing circuitry configured to apply the plurality of weighting values to each of the associated separate gas path parameters from the plurality of expert networks and sum each weighted separate gas path parameter from the plurality of expert networks to a weighted sum value.
2 . The system of claim 1 , wherein the plurality of expert networks each comprises a neural network configured to optimize operation of the gas turbine engine.
3 . The system of claim 1 , wherein the router network further comprises a neural network configured to optimize operation of the gas turbine engine.
4 . The system of claim 1 , wherein the summing circuitry further comprises:
weighting circuitry configured to apply the weighting values to an associated gas path parameter of the plurality of expert networks; and
summer circuitry configured to sum weighted gas path parameters to generate the weighted sum value.
5 . The system of claim 1 further comprising:
a lookup table configured to output a separate gas path parameter responsive to the separate actuator position; and
wherein the summing circuitry is further configured to apply the plurality of weighting values to each of the associated separate gas path parameters from the plurality of expert networks and the lookup table and sum each weighted separate gas path parameter from the plurality of expert networks and the lookup table to a weighted sum value.
6 . The system of claim 5 , wherein the lookup table further comprises a plurality of gas path parameters each associated with a particular separate actuator position.
7 . The system of claim 1 , wherein the separate actuator position comprises controls for at least one of fuel flow, PLA, N1 requested, stator vane angles, bleed emissions.
8 . The system of claim 1 , wherein the at least one ambient condition parameter comprises at least one of Mach number, altitude, ambient temperature, air pressure.
9 . A method for monitoring operation of a gas turbine engine, comprising:
generating a separate gas path parameter responsive to a separate actuator position using a plurality of expert networks;
generating a weighting vector including a plurality of weighting values each associated with one of the plurality of expert networks responsive to at least one ambient condition parameter using a router network;
applying the plurality of weighting values to each of the associated separate gas path parameters from the plurality of expert networks using summing circuitry; and
summing each weighted separate gas path parameter from the plurality of expert networks using the summing circuitry to generate a weighted sum value.
10 . The method of claim 9 , wherein the step of generating the separate gas path parameters for the plurality of expert networks further comprises generating the separate gas path parameters for the plurality of expert networks using a neural network for each of the plurality of expert networks configured to optimize operation of the gas turbine engine.
11 . The method of claim 9 , wherein the step of generating a weighting vector including a plurality of weighting values further comprising generating the weighting vector including the plurality of weighting values using a neural network for the router network configured to optimize operation of the gas turbine engine.
12 . The method of claim 9 , wherein the step of applying further comprises applying the plurality of weighting values to each of the associated separate gas path parameters from the plurality of expert networks and a lookup table configured to output a second separate gas path parameter responsive to the separate actuator position.
13 . The method of claim 12 , wherein the step of summing further comprises summing each weighted separate gas path parameter from the plurality of expert networks and the lookup table to generate a weighted sum value.
14 . The method of claim 12 , wherein the lookup table further comprises a plurality of gas path parameters each associated with a particular separate actuator position.
15 . The method of claim 9 , wherein the separate actuator position comprises controls for at least one of fuel flow, PLA, N1 requested, stator vane angles, bleed emissions.
16 . The method of claim 9 , wherein the at least one ambient condition parameter comprises at least one of Mach number, altitude, ambient temperature, air pressure.
17 . A system for monitoring operation of a gas turbine engine, comprising:
a plurality of expert networks each configured to generate a separate gas path parameter responsive to a separate actuator position, wherein the plurality of expert networks each comprises a first neural network configured to optimize operation of the gas turbine engine;
a router network configured to generate a weighting vector responsive to at least one ambient condition parameter, wherein the router network further comprises a plurality of second neural networks configured to optimize operation of the gas turbine engine, wherein the weighting vector includes a plurality of weighting values each associated with one of the plurality of expert networks;
a lookup table configured to output a separate gas path parameter responsive to the separate actuator position, wherein the lookup table further comprises a plurality of gas path parameters each associated with a particular separate actuator position; and
summing circuitry configured to apply the plurality of weighting values to each of the associated separate gas path parameters from the plurality of expert networks and the lookup table and sum each weighted separate gas path parameter from the plurality of expert networks to generate a weighted sum value.
18 . The system of claim 17 , wherein the summing circuitry further comprises:
weighting circuitry configured to apply the weighting values to an associated gas path parameter of the plurality of expert networks and the lookup table; and
summer circuitry configured to sum weighted gas path parameters to generate the weighted sum value.
19 . The system of claim 17 , wherein the separate actuator position comprises controls for at least one of fuel flow, PLA, N1 requested, stator vane angles, bleed emissions.
20 . The system of claim 17 , wherein the at least one ambient condition parameter comprises at least one of Mach number, altitude, ambient temperature, air pressure.