Aerodynamic parameters from distributed heterogeneous CNT hair sensors with a feedforward neural network

Abstract

Distributed arrays of artificial hair sensors have bio-like sensing capabilities to obtain spatial and temporal surface flow information which is an important aspect of an effective fly-by-feel system. The spatiotemporal surface flow measurement enables further exploration of additional flow features such as flow stagnation, separation, and reattachment points. Due to their inherent robustness and fault tolerant capability, distributed arrays of hair sensors are well equipped to assess the aerodynamic and flow states in adverse conditions. In this paper, a local flow measurement from an array of artificial hair sensors in a wind tunnel experiment is used with a feedforward artificial neural network to predict aerodynamic parameters such as lift coefficient, moment coefficient, free-stream velocity, and angle of attack on an airfoil. We find the prediction error within 6% and 10% for lift and moment coefficients. The error for free-stream velocity and angle of attack were within 0.12 mph and 0.37 degrees. Knowledge of these parameters are key to finding the real time forces and moments which paves the way for effective control design to increase flight agility, stability, and maneuverability.

Distributed Pressure Sensing–Based Flight Control for Small Fixed-Wing Unmanned Aerial Systems

“…An ANN using the signals from all of the pressure ports was able to provide reliable estimates of AoA throughout the range of AoA tested in the wind tunnel (−4°to 16°) with an RMS error practically equal to that of the AoA vane (0.93°v s 0.95°). This level of performance is comparable to other wind tunnel studies where wing mounted arrays of sensors have been used to estimate AoA [6,11,[13][14][15]. The ANN estimate of AoA was comparable to the vane estimate and appeared to provide a similar estimate to the vane post-stall ( Fig.…”

“…The performance of an ANN in estimating airspeed based all of the distributed pressure port information was assessed against the airspeed measurements from a Pitot tube mounted on the wing close to the wing insert. This ANN also had a level of performance comparable to other wind tunnel studies with wing mounted sensors [6,11,[13][14][15]. In outdoor test flights there were occasions when the ANN and Pitot estimates of V differed significantly ( Fig.…”

“…These are similar to the measurements provided by the instruments on a conventional air-data boom, but with the advantage of reduced cost and increased damage tolerance [6]. It has also been demonstrated that wing mounted arrays can be used to estimate non-conventional parameters such as lift, drag and pitching moment [15,16]. These estimates could potentially be used in a physics based control approach where the total forces and moments acting on the aircraft are estimated based on signals from distributed sensors [19].…”

“…A variety of different man-made sensors have also been used to measure the distribution of flow properties on the wings of small UAS. The most common are pressure sensors [6][7][8][9][10][11][12], but more novel sensors such as hotfilms [13,14] and artificial hair sensors [15,16] have also been used to measure flow velocity.…”

High-Bandwidth and Sensitive Air Flow Sensing Based on Resonance Properties of CNT-on-Fiber Hairs

“…Due to the strong resonance, higher-order resonant modes are accessible at even higher frequencies. The fabrication, sensitivity to steady aerodynamics, internal mechanics, and integration of arrays into aerodynamic structures of these CNT-based sensors have been previously discussed [1,4,13]. Essentially, the external aerodynamic loading induces a moment inside the capillary which compresses the radially grown CNT array, providing a sigmoidal resistance change (decrease) as a function of flow velocity or deflection.…”