基于LS-SVM算法和性能可靠性的航空发动机在翼寿命预测方法

Prediction method of aero-engine life on wing based on LS-SVM algorithm and performance reliability

以航空发动机的实际性能监测数据为基础,建立了时变性能退化模型,并进行了性能趋势预测。根据监测数据中大量与在翼寿命紧密相关的信息,分析了性能退化过程与失效分布函数之间的关系,得到了给定可靠度下的航空发动机在翼寿命。以航空发动机的实际在翼寿命为基础,利用K-S拟合检验方法对在翼寿命分布模型进行检验,采用最小二乘支持向量机确定了模型参数。结合性能退化趋势,计算了修正后的航空发动机在翼寿命,并以6台PW4000航空发动机为案例进行实例验证。分析结果表明:当正则化参数分别为25、37、28、40、27与35时,6台PW4000航空发动机的实际在翼寿命依次为6 921、7 160、7 820、8 490、8 498、6 921循环,对应的在翼寿命预测值依次为6 534、6 726、7 378、7 940、9 103、6 534循环,最大相对误差为0.071 190,最小相对误差为0.055 917,平均相对误差为0.060 824,可见,提出的方法可以很好地满足工程实际需要。

Based on the monitoring data of practical performance for aero-engine, the degradation model of time-varying performance was established, and the performance trend was predicted. According to the much information related to aero-engine life on wing in the monitoring data, the relation between the performance degradation process and the failure distribution function was analyzed, and the aero-engine life on wing under a given reliability was obtained. Based on the practical life data on wing for aero-engine, the distribution model of life on wing was tested by using K-S fitting test method, and the model parameters was determined by using least squares- support vector machine （LS-SVM）. Combined with the performance degradation trend, the revised life on wing for aero-engine was calculated, and example verification on six PW4000 aero-engines was carried out. Analysis result shows that when regularization parameters are 25, 37, 28, 40, 27 and 35 respectively, the practical lives on wing for six PW4000 aero-engines are 6 921, 7 160, 7 820, 8 490, 8 498, 6 921 cycles in order, while the corresponding prediction values are 6 534, 6 726, 7 378, 7 940, 9 103, 6 534 cycles in order. The maximum relative error is 0. 071 190, the minimum relative error is 0. 055 917, and the mean relative error is 0.060 824. The practical engineering requirement can be commendably satisfied by using the proposed method. 4 tabs, 16 figs, 25 refs.