Machine learning system for sound exposure level prediction
A method, apparatus, system, and computer program product for predicting a sound exposure level generated by an aircraft. A flight path over a location is identified. The sound exposure level generated by an aircraft for the identified flight path over the location is predicted using a machine learning model trained using a training dataset comprising historical aircraft sensor data for selected parameters, historical atmospheric data, and historical sound data recorded by a microphone system for flight paths over the location.
1 . A method for predicting a sound exposure level generated by an aircraft, the method comprising:
creating a training dataset from historical aircraft sensor data for: selected parameters, historical atmospheric data, and historical sound data recorded by a microphone system, for flight paths over a location, wherein the historical sound data comprises events comprising sound exposure levels, wherein the training dataset comprises groups of data and wherein each group of data in the groups of data is for a flight path in the flight paths and includes the historical aircraft sensor data for the selected parameters, the historical atmospheric data, and the historical sound data recorded by the microphone system corresponding to time instances within an event having the sound exposure level for the flight path, wherein creating the training dataset from the historical aircraft sensor data comprises:
selecting a window of consecutive time instances within an event having the sound exposure level, wherein:
slant distances between the aircraft and a microphone are less than a threshold; and
the selected parameters within the window of the consecutive time instances are stable parameters; having a moving average of each selected parameter in the selected parameters within a moving average threshold; and
correlating the historical aircraft sensor data for the selected parameters, the historical atmospheric data, and the historical sound data for the consecutive time instances in the window with the sound exposure level;
training a set of machine learning models to predict the sound exposure level generated by the aircraft for the flight path over the location using the training dataset; and
predicting the sound exposure level generated by the aircraft for the flight path over the location using the set of machine learning models after training the set of machine learning models using the training dataset.
2 . The method of claim 1 further comprising:
performing feature engineering on the historical aircraft sensor data for the selected parameters, the historical atmospheric data, and the historical sound data recorded by the microphone system to at least one of selecting relevant features, handling missing data, computing additional parameters, normalizing data, standardizing the data, or performing dimensionality reduction.
3 . The method of claim 1 , further comprising:
selecting parameters that have a correlation to the sound exposure level to form correlated parameters; and
removing the correlated parameters that are repetitive to form the selected parameters.
4 . The method of claim 1 , wherein the training dataset comprises groups of data and wherein each group of data in the groups of data is for a flight in flights and includes the historical aircraft sensor data for the selected parameters, the historical atmospheric data, and the historical sound data recorded by the microphone system corresponding to time instances within an event having the sound exposure level for the flight.
5 . The method of claim 1 , further comprising:
repeating, for each microphone in the microphone system that recorded the historical sound data, selecting the window of the consecutive time instances within the event having the sound exposure level, wherein the slant distances between the aircraft and the microphone are less than the threshold and the selected parameters within the window of the consecutive time instances are the stable parameters and correlating the historical aircraft sensor data for the selected parameters, the historical atmospheric data, and the historical sound data for the consecutive time instances in the window with the sound exposure level.
6 . The method of claim 1 , wherein the set of machine learning models comprises a single machine learning model.
7 . The method of claim 1 , wherein the set of machine learning models comprises different machine learning models and further comprising:
identifying a machine learning model from the different machine learning models having a highest level of accuracy in predicting sound exposure levels, wherein the machine learning model having the highest level of accuracy is a selected machine learning model for use in predicting the sound exposure level; and
wherein predicting the sound exposure level for the flight path of the aircraft over the location comprises:
predicting the sound exposure level for the flight path of the aircraft over the location using the selected machine learning model.
8 . The method of claim 7 further comprising:
continuing to train the different machine learning models using new training datasets generated from new aircraft sensor data for the selected parameters, new atmospheric data, and new sound data recorded by the microphone system; and
repeating identifying the machine learning model from the different machine learning models having the highest level of accuracy in predicting the sound exposure levels in response to continuing to train the different machine learning models.
9 . The method of claim 7 further comprising:
adjusting hyperparameters for the different machine learning models to increase accuracy prior to selecting the machine learning model from the different machine learning models having the highest level of accuracy.
10 . The method of claim 7 , wherein the different machine learning models are selected from at least one of a linear machine learning model, a nonlinear machine learning model, an ensemble machine learning model system, or a neural network model.
11 . The method of claim 1 , wherein the flight path is for one of a departing flight from an airport and an arriving flight to the airport.
12 . The method of claim 1 , wherein the training dataset comprises the historical aircraft sensor data for the selected parameters, the historical atmospheric data, and the historical sound data for the flight paths recorded for at least one of an aircraft model, a tail number, or an airline.
13 . The method of claim 1 , wherein the training dataset further comprises airport data.
14 . A method for predicting a sound exposure level, the method comprising:
creating a training dataset from historical aircraft sensor data for: selected parameters, historical atmospheric data, and historical sound data recorded by a microphone system, for flight paths over a location, wherein the historical sound data comprises events comprising sound exposure levels, wherein the training dataset comprises groups of data and wherein each group of data in the groups of data is for a flight path in the flight paths and includes the historical aircraft sensor data for the selected parameters, the historical atmospheric data, and the historical sound data recorded by the microphone system corresponding to time instances within an event having the sound exposure level for the flight path, wherein creating the training dataset from the historical aircraft sensor data comprises:
selecting a window of consecutive time instances within an event having the sound exposure level, wherein:
slant distances between an aircraft and a microphone are less than a threshold; and
the selected parameters within the window of the consecutive time instances are stable parameters; having a moving average of each selected parameter in the selected parameters within a moving average threshold; and
correlating the historical aircraft sensor data for the selected parameters, the historical atmospheric data, and the historical sound data for the consecutive time instances in the window with the sound exposure level;
identifying a flight path over a location; and
predicting the sound exposure level generated by an aircraft for the flight path over the location using a machine learning model trained using a training dataset comprising historical aircraft sensor data for selected parameters, historical atmospheric data, and historical sound data recorded by a microphone system for flight paths over the location.
15 . The method of claim 14 further comprising:
performing a set of actions using a prediction of the sound exposure level generated by the aircraft for the flight path over the location.
16 . The method of claim 15 , wherein the set of actions is selected from at least one of planning a flight over the location using the prediction of the sound exposure level generated by the aircraft for the flight path over the location, determining compliance with a regulation regarding the sound exposure level for the location, changing the flight path, logging the sound exposure level, changing a flight time for using the flight path, or identifying changes to aircraft configuration for the aircraft.
17 . A machine learning system configured to predict sound exposure level prediction, wherein the machine learning system comprises:
a computer system configured to:
create a training dataset from historical aircraft sensor data for: selected parameters, historical atmospheric data, and historical sound data recorded by a microphone system, for flight paths over a location, wherein the historical sound data comprises events comprising sound exposure levels, wherein the training dataset comprises groups of data and wherein each group of data in the groups of data is for a flight path in the flight paths and includes the historical aircraft sensor data for the selected parameters, the historical atmospheric data, and the historical sound data recorded by the microphone system corresponding to time instances within an event having the sound exposure level for the flight path, wherein to create the training dataset from the historical aircraft sensor data the computer system is configured to:
select a window of consecutive time instances within an event having the sound exposure level, wherein:
slant distances between an aircraft and a microphone are less than a threshold; and
the selected parameters within the window of the consecutive time instances are stable parameters; having a moving average of each selected parameter in the selected parameters within a moving average threshold; and
correlate the historical aircraft sensor data for the selected parameters, the historical atmospheric data, and the historical sound data for the consecutive time instances in the window with the sound exposure level; and
a sound exposure level predictor in the computer system, wherein the sound exposure level predictor is configured to:
train a set of machine learning models to predict a sound exposure level generated by an aircraft for a flight path over a location using the training dataset; and
predict the sound exposure level generated by the aircraft for the flight path over the location using the set of machine learning models after training the set of machine learning models using the training dataset.
18 . The machine learning system of claim 17 , wherein the sound exposure level predictor is configured to:
select the selected parameters from parameters that have a correlation to the sound exposure level to form correlated parameters; and
remove the correlated parameters that are repetitive to form the selected parameters.
19 . The machine learning system of claim 17 , wherein the training dataset comprises groups of data and wherein each group of data in the groups of data is for a flight in flights and includes the historical aircraft sensor data for the selected parameters, the historical atmospheric data, and the historical sound data recorded by the microphone system corresponding to time instances for the sound exposure levels for the flight.
20 . The machine learning system of claim 17 , wherein the sound exposure level predictor is configured to:
repeat, for each microphone in the microphone system that recorded the historical sound data, selecting the window of the consecutive time instances within the event having the sound exposure level, wherein the slant distances between the aircraft and the microphone are less than the threshold and the selected parameters within the window of the consecutive time instances are the stable parameters and correlating the historical aircraft sensor data for the selected parameters, the historical atmospheric data, and the historical sound data for the consecutive time instances in the window with the sound exposure level.
21 . The machine learning system of claim 17 , wherein the set of machine learning models comprises a single machine learning model.
22 . The machine learning system of claim 17 , wherein the set of machine learning models comprises different machine learning models and wherein the sound exposure level predictor is configured to:
identify a machine learning model from different machine learning models that comprise a highest level of accuracy in prediction of sound exposure levels, wherein the machine learning model that comprises the highest level of accuracy is a selected machine learning model for use in prediction of the sound exposure level; and wherein the sound exposure level predictor is configured to:
predict the sound exposure level for the flight path of the aircraft over the location using the selected machine learning model.
23 . The machine learning system of claim 22 , wherein the sound exposure level predictor is configured to:
continue to train the different machine learning models using new training datasets generated from new aircraft sensor data for the selected parameters, new atmospheric data, and new sound data recorded by the microphone system; and
repeat identifying the machine learning model from the different machine learning models having the highest level of accuracy in predicting the sound exposure levels in response to continuing to train the different machine learning models.
24 . The machine learning system of claim 22 , wherein the different machine learning models are selected from a linear machine learning model, a nonlinear machine learning model, an ensemble machine learning model system, or a neural network model.
25 . The machine learning system of claim 17 , wherein the flight path is for one of a departing flight from an airport and an arriving flight to the airport.
26 . The machine learning system of claim 17 , wherein the training dataset comprises the historical aircraft sensor data for the selected parameters, the historical atmospheric data, and the historical sound data for the flight paths recorded for at least one of an aircraft model, a tail number, or an airline.
27 . The machine learning system of claim 17 , wherein the training dataset further comprises airport data.
28 . A computer program product configured to predict a sound exposure level generated by an aircraft, wherein the computer program product comprises a computer readable storage medium that comprises program instructions embodied therewith, wherein the program instructions are configured for execution by a computer system to perform a method of:
creating a training dataset from historical aircraft sensor data for: selected parameters, historical atmospheric data, and historical sound data recorded by a microphone system, for flight paths over a location, wherein the historical sound data comprises events comprising sound exposure levels, wherein the training dataset comprises groups of data and wherein each group of data in the groups of data is for a flight path in the flight paths and includes the historical aircraft sensor data for the selected parameters, the historical atmospheric data, and the historical sound data recorded by the microphone system corresponding to time instances within an event having the sound exposure level for the flight path, wherein creating the training dataset from the historical aircraft sensor data comprises:
selecting a window of consecutive time instances within an event having the sound exposure level, wherein:
slant distances between the aircraft and a microphone are less than a threshold; and
the selected parameters within the window of the consecutive time instances are stable parameters; having a moving average of each selected parameter in the selected parameters within a moving average threshold; and
correlating the historical aircraft sensor data for the selected parameters, the historical atmospheric data, and the historical sound data for the consecutive time instances in the window with the sound exposure level;
identifying a flight path over a location; and
predicting the sound exposure level generated by the aircraft for the flight path over the location using a machine learning model trained using the training dataset comprising historical aircraft sensor data for selected parameters, historical atmospheric data, and historical sound data recorded by a microphone system for flight paths over the location.