Methods for optimized engine balancing based on flight data

Patent metadata

Field	Value
Publication number	US-10343784-B2
Application number	US-201715616589-A
Country	US
Kind code	B2
Filing date	Jun 7, 2017
Priority date	Jun 7, 2017
Publication date	Jul 9, 2019
Grant date	Jul 9, 2019

Abstract

Methods for balancing aircraft turbofan engines using an artificial neural network and a global optimization algorithm, which are configured as a loop in which: the global optimization algorithm inputs to the artificial neural network data representing proposed changes in the balance weight attached to the engine; the artificial neural network outputs data estimating changes in engine vibration in response to input of the data representing the proposed changes in balance weight; and the global optimization algorithm outputs an optimum influence coefficient calculated using a minimum predicted change in engine vibration. This process can be repeated multiple times to derive a multiplicity of optimum influence coefficients which are used to determine an optimal balance solution for the engine. The balance weights attached to the engine are then reconfigured in accordance with the optimal balance solution.

First claim

The invention claimed is:

1. A method for balancing an unbalanced gas turbine aircraft engine, comprising:
(a) acquiring flight parameter data, balance weight data and vibration data for a multiplicity of flights of an aircraft having a gas turbine engine that is unbalanced;
(b) using the flight parameter data, balance weight data and vibration data to train an artificial neural network to output difference (Δ) vibration data representing a predicted Δ vibration for the gas turbine engine for first and second flights of the aircraft in response to input of flight parameter data for the first and second flights and difference (Δ) balance weight data representing a Δ balance weight, wherein the predicted Δ vibration corresponds to a difference between a predicted engine vibration for the second flight and a measured engine vibration for the first flight, while the Δ balance weight corresponds to a difference between the respective balance weights for the first and second flights;
(c) iteratively performing a loop of operations comprising using the artificial neural network and a global optimization method to calculate a first optimum influence coefficient, wherein the global optimization method determines a first minimum predicted Δ vibration output by the artificial neural network for modeled flights having various changed Δ balance weights, wherein the first minimum predicted Δ vibration is for engine revolutions per minute (rpm) in a first engine rpm range and for altitudes in a first altitude range;
(d) determining an optimal balance solution for the gas turbine engine using the first optimum influence coefficient calculated in step (c); and
(e) balancing the gas turbine engine by attaching to or removing from the gas turbine engine one or more balance weights each having a respective mass and a respective location in accordance with the optimal balance solution;
wherein in step (b) the artificial neural network is trained to output Δ vibration data representing a difference in amplitude and a difference in phase of predicted vibrations in the gas turbine engine during the first and second flights of the aircraft in response to input of flight parameter data and Δ balance weight data for the first and second flights;
wherein the flight parameter data comprises data representing engine rpm, altitude, Mach number, temperature and thrust for the first and second flights, and the Δ balance weight data represents real and imaginary components of the Δ balance weight for the first and second flights; and
wherein the global optimization method comprises:
inputting to the artificial neural network Δ balance weight data representing various changed Δ balance weights; and
determining the first minimum predicted Δ vibration in the gas turbine engine based on the Δ vibration data that is output from the artificial neural network in response to inputting to the artificial neural network of Δ balance weight data representing various changed Δ balance weights.

2. A method for balancing an unbalanced gas turbine aircraft engine, comprising:
(a) acquiring flight parameter data, balance weight data and vibration data for a multiplicity of flights of an aircraft having a gas turbine engine that is unbalanced;
(b) using the flight parameter data, balance weight data and vibration data to train an artificial neural network to output difference (Δ) vibration data representing a predicted Δ vibration for the gas turbine engine for first and second flights of the aircraft in response to input of flight parameter data for the first and second flights and difference (Δ) balance weight data representing a Δ balance weight, wherein the predicted Δ vibration corresponds to a difference between a predicted engine vibration for the second flight and a measured engine vibration for the first flight, while the Δ balance weight corresponds to a difference between the respective balance weights for the first and second flights;
(c) iteratively performing a loop of operations comprising using the artificial neural network and a global optimization method to calculate a first optimum influence coefficient, wherein the global optimization method determines a first minimum predicted Δ vibration output by the artificial neural network for modeled flights having various changed Δ balance weights, wherein the first minimum predicted Δ vibration is for engine revolutions per minute (rpm) in a first engine rpm range and for altitudes in a first altitude range;
(d) determining an optimal balance solution for the gas turbine engine using the first optimum influence coefficient calculated in step (c); and
(e) balancing the gas turbine engine by attaching to or removing from the gas turbine engine one or more balance weights each having a respective mass and a respective location in accordance with the optimal balance solution,
wherein the first optimum influence coefficient is calculated by dividing a modeled Δ balance weight that resulted in the first minimum predicted Δ vibration by the first minimum predicted Δ vibration.

3. A method for balancing an unbalanced gas turbine aircraft engine, comprising:
(a) acquiring flight parameter data, balance weight data and vibration data for a multiplicity of flights of an aircraft having a gas turbine engine that is unbalanced;
(b) using the flight parameter data, balance weight data and vibration data to train an artificial neural network to output difference (Δ) vibration data representing a predicted Δ vibration for the gas turbine engine for first and second flights of the aircraft in response to input of flight parameter data for the first and second flights and difference (Δ) balance weight data representing a Δ balance weight, wherein the predicted Δ vibration corresponds to a difference between a predicted engine vibration for the second flight and a measured engine vibration for the first flight, while the Δ balance weight corresponds to a difference between the respective balance weights for the first and second flights;
(c) iteratively performing a loop of operations comprising using the artificial neural network and a global optimization method to calculate a first optimum influence coefficient, wherein the global optimization method determines a first minimum predicted Δ vibration output by the artificial neural network for modeled flights having various changed Δ balance weights, wherein the first minimum predicted Δ vibration is for engine revolutions per minute (rpm) in a first engine rpm range and for altitudes in a first altitude range;
(d) determining an optimal balance solution for the gas turbine engine using the first optimum influence coefficient calculated in step (c);
(e) balancing the gas turbine engine by attaching to or removing from the gas turbine engine one or more balance weights each having a respective mass and a respective location in accordance with the optimal balance solution;
(f) iteratively performing a loop of operations comprising using the artificial neural network to generate predicted Δ vibration data representing respective predicted Δ vibrations for respective pairs of flights of the aircraft for engine rpm in the first engine rpm range and altitudes in the first altitude range, and then calculating a set of respective influence coefficients based on the respective predicted Δ vibration data and respective Δ balance weights for the respective pairs of flights; and
(g) calculating a statistical measure for the set of influence coefficients calculated in step (f).

4. The method as recited in claim 3 , wherein the statistical measure is an arithmetic mean.

5. The method as recited in claim 4 , wherein step (d) also uses the statistical measure of the set of influence coefficients calculated in step (f) to determine the optimal balance solution.

Assignees

• Boeing Co

Inventors

• Nakhjavani Omid B

Classifications

• G06N3/08

  Learning methods · CPC title

• G01M1/36

  by adjusting position of masses built-in the body to be tested · CPC title

• F01D17/02

  Arrangement of sensing elements · CPC title

• B64D2045/0085

  Devices for aircraft health monitoring, e.g. monitoring flutter or vibration · CPC title

• B64D45/00

  Aircraft indicators or protectors not otherwise provided for · CPC title