Prediction of Peak Stress in Blades from Multi-Mode Experimental Data

The prediction of the oscillatory stresses distribution in blades in their operating conditions is a critical issue in the estimation of their fatigue life. Experimental data to substantiate this prediction is limited due to the large cost of conducting the needed tests but also due to instrumentation challenges. For example, strain gages generally require external wiring passing through a slip ring. Further, tip timing or Non-intrusive Stress Measurement System (NSMS) measurements can also be used but they provide limited information, i.e. the response at one or two selected points at the blade tip. Given the cost of the tests, it is generally desired to capture the stresses in a variety of operating conditions in which different mode shapes are excited using the same, few (one to three say) strain gages and/or tip timing measurements. To observe the expected vibratory modes of interest, of number larger than the number of actual sensors, and to minimize the effects of mispositioning, the gages are typically not placed at the expected locations of the peak stresses but gage factors are given that relate the peak strain to the strain measured at the gages and/or the deflection measured at the tip for each mode of interest. The extraction of the stress distribution, in particular the estimation of the peak stress, from this limited data has usually been achieved under the assumption that only a single mode is excited at the operating condition selected. Then, the peak stress on the blade can be recovered by multiplying the measured strain by the gage factor. When more than one mode is significantly present in the response however, this approach can still be used but only to provide an upper bound to the peak stress as the locations at which the peak stress is accomplished in different modes is in general different. To recover the true peak stress, it is necessary to know the stress distribution in more detail as would be obtained from a finite element model but this model is generally not available to the user. Further, the construction of detailed finite element model is a time consuming tasks requiring the measurement of the exact geometry and thus this option is not appropriate for the present effort.