Optimization of selecting strain measurement locations for distributed load recovery from strain measurements

icaf2023 Tracking Number 3

In order to solve the reverse problem of structural load distribution recovery from strain measurements, this paper focuses on how to select strain measurement locations in the influence coefficient method on the basis of the previous research. An optimization procedure of selecting strain measurement locations based on basis strains selection method is proposed. It is matched with the Euclidean space method based on Schmidt's orthogonalization of the maximum vertical distance proposed in Ref[1,2]. The optimization steps are as follows: a) Construct load column space according to design load cases, and stepwise select load basis cases based on Schmidt's orthogonalization; b) Construct strain row space according to candidate strain set, and stepwise select basis strains based on Schmidt's orthogonalization; c) Construct strain column space according to basis strains matrix, and stepwise select strain basis cases based on Schmidt's orthogonalization, then influence coefficient matrix [A] is determined; d) Calculate the determinant of [A]T[A]; e) Change the initial basis strain in step b), repeat step b) ~d), and finally select [A] with the maximum det([A]T[A]), then the optimal influence coefficient matrix and basis cases are determined. Taking the load rams and optical fiber sensor data in a certain full-scale wing fatigue test as a case study, the feasibility of the optimization procedure of selecting strain measurement locations is verified under the condition that the number of strain measurements is limited. The load prediction accuracy under different numbers of strain measurements and different strain measurement errors is compared and analysed. It is verified that basis strains selection method combined with the load distribution recovery based on Euclidean space method can provide load predictions of quite high accuracy and robustness. The load prediction accuracy is higher than traditional load calibration program based on strain bridges. With the decrease of the number of strain measurements, the precision of load recovery is reduced. It is suggested that the number of strain measurements should be at least twice the number of basis load cases in practical application.