The existing flush air data sensing systems have some deficiencies such as singularity values in calculat-ing air data. Hence we propose a flush air data sensing algorithm based on the pseudo-inverse matrix and back-propagation ( BP ) neural networks, which we believe can overcome the deficiencies. The core of the algorithm con-sists of:(1) it uses the three-point method to estimate the local angle of attack and sideslip of an aircraft and diag-nose its faults at pressure points;it then uses the pseudo-inverse matrix with fault tolerance to solve the total pres-sure and amend the dynamic pressure;(2) it utilizes the strong nonlinear mapping capability of the BP neural net-works to fit the nonlinear mathematical model of the flush air-data sensing system, thus reducing the number of di-mensions of input vectors and the level of difficulty in training networks and achieving the measurement calibration. The simulation results, given in Tables 1 and 2, and their analysis show preliminarily that our new algorithm has better fault tolerance and can produce highly precise and reliable air data.%针对现有FADS算法存在的不足,提出了一种融合广义逆和BP神经网络的高精度嵌入式大气数据传感系统算法。该算法的特点是:①应用三点法预估当地迎角和当地侧滑角,并对测压点进行故障诊断;然后用具有容错能力的广义逆矩阵求解总压力和修正动压;②应用BP神经网络具有的强大非线性映射能力,拟合FADS系统的非线性数学模型,减少输入向量的维数和网络训练难度,完成测量校正。结果表明,所提出的FADS算法在精度、可靠性等方面均有较好的性能。
展开▼