论文部分内容阅读
讨论了具有交叉敏感的传感器组成的多传感器系统的动态特性及其对系统检测精度的影响. 提出了将该多传感器系统看作一个线性滤波器与一个无记忆非线性函数的串联系统, 即Wiener系统, 把后续的信息融合系统看作一个无记忆非线性函数与一个线性滤波器的串连, 即Hammerstein系统. 在静态标定的基础上, 用盲解卷积技术求得Hammerstein的线性滤波器系数矩阵. 针对盲解卷积技术复现信号幅值不确定性的缺点, 提出了一种根据逆滤波器系数矩阵与静态标定的线性矩阵的同一性校正滤波器系数矩阵的方法, 以获得多传感器系统的近似逆向动态模型, 提高了多传感器系统在实际检测过程中的检测精度, 降低了动态检测结果失真度, 满足了实际情况的需要. 最后的仿真结果表明, 输入信号频率接近采样频率1/10时, 用盲解卷积技术获得的幅值复现误差是未经过动态补偿的幅值复现误差的1/20, 是各传感器单独动态补偿后幅值复现误差的1/2, 快速性提高了2倍. 金属氧化物半导体甲烷传感器的动态补偿结果表明, 在温度阶跃响应下, 补偿后的甲烷检测误差小于补偿前的1/2, 因此这种方法拓宽了多传感器系统的带宽.
The dynamic characteristics of multisensor system with cross-sensitive sensors and its influence on the system’s detection accuracy are discussed.The multisensor system is proposed as a series system with a linear filter and a memoryless nonlinear function, ie Wiener System, the subsequent information fusion system is regarded as a series connection of a memoryless nonlinear function and a linear filter, ie Hammerstein system.Based on the static calibration, the Hammerstein linear filter coefficients are obtained by blind deconvolution Matrix.Aiming at the shortcomings of the signal amplitude uncertainty of the blind deconvolution technique, a method of correcting the coefficient matrix of the filter according to the identity of the linear matrix of the static filter and the inverse filter coefficient matrix is proposed to obtain the multi-sensor System approximate inverse dynamic model to improve the detection accuracy of the multisensor system in the actual detection process and reduce the distortion of the dynamic detection results to meet the needs of the actual situation.The final simulation results show that the input signal frequency close to the sampling frequency 1 / At 10 o’clock, the amplitude reproducibility error obtained with the blind deconvolution technique is not dynamically compensated 1/20 of the amplitude reproducibility error is 1/2 of the amplitude reproducibility error of each sensor after dynamic compensation alone and the rapidity is improved by 2 times.The dynamic compensation results of the metal oxide semiconductor methane sensor show that in the temperature step In response, the compensated methane detection error is less than ½ prior to compensation, so this approach broadens the bandwidth of the multisensor system.