随机振动分析中遇到的一个很诡异的错误

hunanyangdong

我做的一个大坝的随机振动，只考虑行波效应，相干系数为1，在进行模态提取、扩展后，进行谱分析时出现问题，小弟困惑不已，还望路过的高手不吝赐教！

** ERROR **                           CP =     342.766   TIME= 10:36:29
The co- or quad- spectrum for table numbers 1 and 2 at the frequency   
0.7958 does not satisfy the coherence to be between 0 and 1.   
其实从我的互谱矩阵只考虑行波效应，相干系数为1，
PSDVAL,1, 0.00151334,0.00183134,0.00245431,0.00273046,0.00221413,0.00153141,0.00104741
PSDVAL,1, 0.00074342,0.00055051,0.00042303,0.00033510,0.00027209,0.00022544,0.00018994
PSDVAL,1, 0.00016229,0.00014033,0.00012259,0.00010805,0.00009597,0.00008583,0.00007723
PSDVAL,1, 0.00006987,0.00006353,0.00005801,0.00005319,0.00004895,0.00004520,0.00004187
PSDVAL,1, 0.00003890,0.00003623,0.00003383,0.00003166,0.00002970,0.00002791,0.00002628
PSDVAL,1, 0.00002479,0.00002342,0.00002217,0.00002101,0.00001994,0.00001895,0.00001804
PSDVAL,1, 0.00001718,0.00001639,0.00001565,0.00001496,0.00001432,0.00001371,0.00001315
PSDVAL,2, 0.00151334,0.00183134,0.00245431,0.00273046,0.00221413,0.00153141,0.00104741
PSDVAL,2, 0.00074342,0.00055051,0.00042303,0.00033510,0.00027209,0.00022544,0.00018994
PSDVAL,2, 0.00016229,0.00014033,0.00012259,0.00010805,0.00009597,0.00008583,0.00007723
PSDVAL,2, 0.00006987,0.00006353,0.00005801,0.00005319,0.00004895,0.00004520,0.00004187
PSDVAL,2, 0.00003890,0.00003623,0.00003383,0.00003166,0.00002970,0.00002791,0.00002628
PSDVAL,2, 0.00002479,0.00002342,0.00002217,0.00002101,0.00001994,0.00001895,0.00001804
PSDVAL,2, 0.00001718,0.00001639,0.00001565,0.00001496,0.00001432,0.00001371,0.00001315

COVAL,1,2, 0.00132808,-.00146717,-.00051736,0.00266653,-.00133293,-.00072824,0.00104511
COVAL,1,2, -.00026386,-.00038668,0.00039745,-.00002666,-.00023762,0.00018180,0.00003839
COVAL,1,2, -.00015818,0.00008547,0.00005734,-.00010787,0.00003486,0.00005974,-.00007279
COVAL,1,2,0.00000618,0.00005520,-.00004708,-.00001029,0.00004761,-.00002785,-.00001926
COVAL,1,2, 0.00003885,-.00001346,-.00002333,0.00002994,-.00000289,-.00002413,0.00002147
COVAL,1,2, 0.00000458,-.00002273,0.00001381,0.00000950,-.00001993,0.00000719,0.00001232
COVAL,1,2, -.00001630,0.00000174,0.00001346,-.00001230,-.00000252,0.00001328,-.00000828
QDVAL,1,2, -.00072553, -.00109601, 0.00239916, -.00058738, -.00176796, 0.00134718, 0.00006947
QDVAL,1,2, -.00069502, 0.00039184, 0.00014488, -.00033404, 0.00013256, 0.00013332, -.00018602
QDVAL,1,2, 0.00003631, 0.00011130, -.00010836, -.00000621, 0.00008942, -.00006162, -.00002581
QDVAL,1,2, 0.00006960, -.00003144, -.00003389, 0.00005219, -.00001138, -.00003561, 0.00003718
QDVAL,1,2, 0.00000189, -.00003364, 0.00002450, 0.00001032, -.00002956, 0.00001403, 0.00001516
QDVAL,1,2, -.00002436, 0.00000564, 0.00001734, -.00001874, -.00000079, 0.00001753, -.00001317
QDVAL,1,2, -.00000545, 0.00001630, -.00000799, -.00000852, 0.00001409, -.00000342, -.00001021

从如上数据也可以看出虚部和实部的平方和等于自谱的平方
即(-.00146717)^2+( -.00109601) ^2=0.00183134^2

而当我把0.79580Hz处的谱值全部用0.15920Hz的代替后，又出现下面的错误提示
The co- or quad- spectrum for table numbers 1 and 2 at the frequency   
2.70560 does not satisfy the coherence to be between 0 and 1.  
请问互功率谱的定义时，小数位数有限制吗？

hunanyangdong

我的自功率谱是用金井清公式编程序算出来的，然后只考虑行波效应，即相干系数为1，互功率谱项为S*EXP(-iwt)，然后为了得到实部COVAL和虚部QDVAL，我使用欧拉公式，即EXP(-iwt)=cos(wt)-i*sin(wt)，这样实部便为S*cos(wt)，
虚部便为S*sin(wt)，由于保留8位小数，导致实部和虚部的平方和与自谱的平方和出现偏差，不知道如何处理是好？