We use CAESAR II to do dynamic anlysis,we set the cutoff freq as 33Hz,we get the result as follow:
1220 0 0 259117 0 0 0 Rigid +Z
160913
1 Z/DM
For the same pipe and same restraint,we set the cutoff freq as 100Hz,we get the smaller result as follow:
1220 0 0 226254 0 0 0 Rigid +Z
133770
1 Z/DM
We can not understand why more modes considered will get the smaller result,which cutoff freq shall we select?
Thanks for your kind reply in advance.

How is your acceleration by 33 Hz ?
Have you acceleration input > 33 HZ ?
Yes ? How much ?

Although there may be exceptions, what you show makes sense to me. In both cases the "missing mass component" provides more response than any individual mode of vibration. (That's the M in 1 Z/DM but I do not recognize the D in that string.)
As you extend the frequency analysis, more modes of vibration are included in the modal evaluation and the remainder of the system that is not included in the dynamic evaluation - the missing mass component - is reduced. If your load continues to drop as frequency increases, the load factor used for the missing mass component will be lower at the higher frequency cutoff. So teh missing mass response and total response drops. That's what I see in your brief output report.

I would consider your second analysis as having less "round-off" and, therefore, more accurate.

Thanks above.
Why we choose 100Hz as cutoff freq,because there is a spectrum's cutoff freq is 100Hz,but most spectrum's cutoff freq is 33Hz.
Out model is main steam pipe,and we put all steam hammer force in one case.

33Hz is a reasonable cutoff for seismic analysis. Earth motion is not "fast enough" to cause resonant response of system modes of vibration greater than about 33Hz. Hammer loads, however, may produce resonant response at higher frequencies.

Thanks Dave's reply,I check my result doc again,and find some rules as follows:
1.The big different value exist in rigid restraint,refer to snubber,the load of 100Hz is as same as 33Hz.
2.If we set gap in rigid restraint,in 100Hz,the load is about 600000N,in 33Hz,the load is 800000N.
According to my experience,more rigid mode input will lead to more load,especially load in rigid restraint,but above result is not as so.