基于瞬时转速统计分析的内燃机失火检测

Misfire Detection Based on Statistical Analysis of Instantaneous Rational Speed

多缸内燃机的瞬时转速序列经过长度调整、滤波以及切分和重组,得到瞬时转速异常波动信号的数据矩阵,从而构造了适合于主分量分析的多变量统计分析问题。利用第一主分量与矩阵各行的相关系数定义了衡量气缸健康状态的特征参数,并以特征参数的均值和标准差自适应生成了用于失火检测的阈值。通过第一主分量的极性判断和气缸特征参数的阈值检验,实现了失火气缸的定位。试验结果证明了这种失火检测方法的鲁棒性。即使在高速轻载条件下,该方法也能识别多缸失火故障,并且所得特征参数可直接用于评价气缸做功的均匀性。

This paper presents a novel method to detect the misfire in internal combustion engines. This method is based on principal component analysis （PCA） of abnormal fluctuation signal in instantaneous rotational speed （IRS）. Abnormal fluctuation signal is acquired by filtering out the firing frequency and its harmonic components. Before filtering, the length of IRS series is adjusted within one engine cycle if necessary. Then, the abnormal fluctuation signal is split and reassembled according to each cylinder, yielding a data matrix. This process converts the single variable analysis into muhivariable analysis, in which PCA can be applied. A dimensionless feature which reflects the working ability of each cylinder is defined with the correlation coefficient between the first principal component （PC） and the corresponding row vector in the matrix above. And the threshold is calculated for each engine cycle adaptively. By chec- king the polarity of the first PC, the malfunction cylinder is identified. Experimental results in a six-cylin- der diesel engine show that this method can detect muhiple misfire even under high speed and low loads condition. And the dimensionless feature can also be used to directly evaluate the uniformity of engine torque.