A workload monitoring system that calculates neck forces in fast jet pilots.

Research output: Contribution to conference (non-published works)Poster

Abstract

PURPOSE: Neck pain in fast jet pilots is an international concern. Helmet design, head checks and high G maneuvers are thought to contribute. In order to understand neck pain and neck forces in fast jet pilots, this study sought to build a model to predict the forces acting at the neck and evaluate the efficiency and accuracy of the subsequent algorithm using only head co-ordinates and Gz flight data from helmet mounted inertial sensors.

METHODS:
Lab derived model:
Three-dimensional motion data was collected using the VICON camera system from RAAF fast jet pilots performing common “head checks” whilst seated on an F/A-18A ejection seat. Pilots were modelled wearing 2 helmet types (HGU-55/P and JHMCS). Dynamic graphical models were derived using Opensim software. The OpenSim model calculated forces at each cervical segment for 1-9Gz conditions.
Three ensemble learning algorithms- linear regression, k- nearest neighbours (kNN) and adaptive boosting were deployed and trained 5 times on random stratified samples of 75% of the Opensim model dataset to establish an algorithm that could predict segment moments at C1, C4 and C7 using only head co-ordinates and Gz. The resultant algorithms were tested on the remaining 25% of the OpenSim dataset and the best model identified based on predictive accuracy and efficiency.
Real time flight model:
Head position co-ordinates and Gz values were obtained from helmet-mounted sensors during 42 sorties involving 7 pilots over 7 days. Unfiltered raw data from flight was analysed by the most accurate machine learner and the resultant predicted cervical spine moments collated.

RESULTS: The Adaptive boosting learner performed best (R2=0.994, RMSE=1.129), kNN (R2=0.981, RMSE 2.013), linear regression was least accurate (R2=0.533, RMSE= 9.953). Predictive errors were in the range 0.7-3%. The kNN algorithm was selected for predicting the real flight data to balance efficiency and accuracy. 42 samples from sorties totalling 223 minutes of flight were analysed successfully by the machine learner. Predicted moments at 3 cervical segments, duration, frequency and direction were determined for all flights in less than 5 minutes.

CONCLUSION: The results demonstrate that neck forces can be measured in dynamic flight environments. Helmet design, Gz and pilot head posture each influence magnitude of neck forces.

OPERATIONAL RELEVANCE: From this system both neck muscle and joint loads can be determined, within a maneuver, within a sortie, within a training week or longer. This technology will enable instructors and health staff to quantify neck workload of pilots and review technique in order to better understand flight related neck pain. Helmet design can also be modelled using this system.
Original languageEnglish
Pages1-28
Number of pages28
Publication statusPublished - 2019
EventDefence Human Sciences Symposium 2019: DHSS 2019 - Canberra, Canberra, Australia
Duration: 19 Nov 201920 Nov 2019
https://www.dst.defence.gov.au/event/defence-human-sciences-symposium-2019

Other

OtherDefence Human Sciences Symposium 2019
Abbreviated titleDHSS 2019
CountryAustralia
CityCanberra
Period19/11/1920/11/19
Internet address

Fingerprint Dive into the research topics of 'A workload monitoring system that calculates neck forces in fast jet pilots.'. Together they form a unique fingerprint.

  • Cite this

    Newman, P., & Spratford, W. (2019). A workload monitoring system that calculates neck forces in fast jet pilots.. 1-28. Poster session presented at Defence Human Sciences Symposium 2019, Canberra, Australia.