Goal

Develop an empirical emission model for aircraft auralisation

Earlier work

• EuroNoise 2015 and VASTCON June 9th, 2015
• Determining an empirical emission model for the auralization of jet aircraft
  • Backpropagation
  • Extracting features
    • Harmonics using complex cepstrum

Overview

• What are the model requirements?
• How to obtain model features?
• Are the extracted features sufficient to describe the emission?

What are the model requirements?

• Predict emission of several aircraft/engine models
• Additive synthesis
• Tones and bandpass-filtered noise
• Directivity for each component

How to obtain model features?

• Extract features as function of emission angles
  • Tones, noise bands
• Backpropagate to source in time-domain
  • Undo geometrical spreading, Doppler shift, atmospheric attenuation
  • Not ground reflection
• Relate features to velocity and engine shaft frequency N₁

Recording

Backpropagated recording

Fan tones

• Buzz-Saw tones are multiples of N₁
• Engine shaft and fan rotate at N₁ frequency
• Blade passing frequency f_bpf = N₁ ⋅ n_blades
• Strong directivity

Front of aircraft - fan noise

Rear of aircraft - jet noise

Determine fundamental frequency

• Determine centerfrequencies of tones
  • Tone-seeking algorithm of ISO 1996-2:2007 Annex C
  • Detects some tones, not all
• Assume all observed tones are harmonics f_n
  • f_n/f₀ = n
  • A priori knowledge of range f₀
• Estimate fundamental

Extract features tonal components

• Assess tones (~ ISO 1996-2:2007 Annex C)
• Centerfrequency
• Bandwidth -3 dB
• Power

Noise components

• Frequency lines that are not part of tones are noise
• 1/3-octaves

Are the extracted features sufficient to describe the emission?

Synthesise emission

• Generate emission for specific event using features
• Phase relation tones?
  • Buzz-Saw tones
• No directivity function
• Emission for reflection same as for direct contribution

Emission synthesis

Immission synthesis

Preliminary listening test

• Group listening test
  • 11 participants, headphones
• Part A - individual stimuli
  • Which stimuli is 'real'?
  • Annoyance rating?
• Part B - paired comparison, same event
  • Which stimuli is 'real'?
  • Similarity rating?
• Part C
  • Questions, comments

Preliminary results

• 80% of recordings were identified as such
• 20% of auralisations were identified as recordings
• 85% was identified correctly in direct comparison
• Annoyance rating auralisations slightly higher
  • Auralisations rough, metallic
  • Cause: phase relation harmonics

Conclusions

• Development of an emission model
• Inverse propagation model
• Automated procedure for extracting features
• Emission synthesis
• Preliminary listening test
• Improvements needed

Thank you for your attention!

The research leading to these results has received funding from the People Programme (Marie Curie Actions) of the European Union's Seventh Framework Programme FP7/2007-2013 under REA grant agreement number 290110, SONORUS "Urban Sound Planner".

Extra

Future work

• Improve emission synthesis
  • Phase distortion?
  • Blade passing frequency
• Extract features from all events
  • ~2600 events and 7 recordings/event
• Statistical analysis, extract directivities
• Listening tests

Emission comparison

Immission comparison

Atmospheric attenuation

• Attenuation coefficient α calculated according to ISO 9613-1
  • Parameters from measurements
• Inverse DFT of attenuation spectrum results in impulse response
• M amount of impulse responses determined for an M amount of equispaced ranges.
• Time-variant convolution

Geometrical spreading

• Amplitude scaling
• Propagation delay
  • Movement results in Doppler shift
  • Shift samples
  • Delay is non-integer amount of samples
  • Lanczos interpolation

Ground reflection?

• Transfer function possible?
  • Phase sensitivity
  • Directivity
• Ignored
• Inaccuracies
  • Estimation of power (interference)
  • Broadening or extra tones (Doppler shift, directivity)