Optimum Sensor Placement for Impact Location Using Trilateration

A key problem associated with structural health monitoring (SHM) is the placement of sensors upon a structure to detect the existence, location, and the extent of any damage. Because input data coming from the sensors are groups of measurements, it is arguable that the most widely used approach to SHM nowadays is to consider it as a statistical pattern recognition problem. Artificial neural networks have made a great impact on pattern recognition practice. A problem associated with this monitoring strategy is to find a good compromise between the quality of information achieved by the sensor network, increasing with the sensor density, and the need to keep the minimum weight and instrumentation cost. Thus, the number of sensors must be kept under control, and a search of the optimal location of such sensors needs to be performed. All these aspects have been taken into account in the present work, dealing with the problem of optimum sensor placement for impact location on a multilayered composite structure. Multilayered composite structures may suffer particularly relevant trauma when subject to low-velocity impacts, as they may produce non-visible or barely visible damage on the structure surface, while remarkable subsurface delaminations may be present. Such hidden damage, when remaining undetected, may grow to catastrophic failure. To overcome this issue, a neural network approach has been used here to predict the impact locations on a composite panel from time-dependent data recorded on a set of surface-mounted piezoelectric sensors during an experimental impact test. A genetic algorithm has been used to find the optimal sensor layout that minimised the error in predicting the impact location. A new approach, based on trilateration, is discussed and compared with the traditional one and is shown to provide the same degree of accuracy at reduced computational cost.