Multi-Objective structural optimization of repairs of blisk blades

Abstract

Modern manufacturing technologies offer multiple options to extend the service life of expensive jet engine components through repairs. In this context, the repair processes of blade-integrated disks (blisks) are of particular interest, as the complex design makes replacement of this part very costly. However, currently, repairs of blisks are mainly done manually and repair design decisions still rely on the expertise of maintenance technicians. From a scientific perspective, these subjective, experience-based decisions are a major drawback, as today’s computational methods allow for systematic analysis and evaluation of design alternatives. The present doctoral thesis contributes to the decision-making process related to the repair of blisk blades by blending and patching by providing an engineering optimization framework and simulation routines for structural assessment of different repair designs. First, an object-oriented optimization framework is developed that is ideally suited to address engineering optimization problems such as blisk repair optimization. The design of the software architecture is chosen to achieve a high degree of flexibility and modularity. In particular, the framework provides a unified interface for global and local derivative-free optimization algorithms and custom engineering optimization problems. Thereby, optimization of single- as well as multi-objective problems is supported. The broad applicability of the framework in engineering optimization is demonstrated using examples from wind energy research. Furthermore, the optimization framework forms a suitable environment for structural multi-objective optimization of blend and patch repairs. The second part of this thesis is devoted to the application of the optimization framework to blend repairs of a compressor blisk. The geometry of the removed blade part and the resulting blend is parameterized by three geometric design variables. The two objectives of the optimization correspond to two modal criteria, because especially the vibration behavior of blades is affected by this kind of geometric modification. To check if frequency requirements are harmed by the repair the first objective reflects the deviation of the natural frequencies of the repaired blade to the natural frequencies of the nominal blade. The second objective considers resonance conditions by evaluating the proximity of natural frequencies to excitation frequencies. Pareto optimal repair designs are found by solving the derived optimization problem using appropriate structural mechanics models of a blade sector and employing the developed optimization framework. By analyzing the optimal blend shapes for two different damage patterns, it is shown that the characteristics of Pareto frontiers, like the occurrence of discontinuities, are damage-specific. Therefore, it is concluded that design decisions on blend repairs have to be made on a case-by-case basis. The third part of this thesis is concerned with the multi-objective optimization of patch repairs. While blend repairs change the blade geometry, patch repairs restore the original blade contour. In terms of structural integrity, the most significant modification due to patching is hence associated with the welding process to join patch and blade. The remaining residual stresses, affect the strength of the repaired blade, are therefore the most critical aspect of patch repairs. Utilizing the engineering optimization framework and the parametric simulation model, a multi-objective optimization problem is solved considering the length of the weld and the fatigue strength of the repaired blade. In addition to fatigue strength properties, the weld length is selected as an optimization goal, since the manufacturing effort of the high-tech repair is of practical importance. Pareto optimal repair designs are presented for a damage pattern at the leading edge. The optimization results are further complemented by subsequent thermal and mechanical simulations of the welding and heat treatment process. Different patch geometries are classified from the Pareto optimal solutions. Depending on the preferences in terms of weld length and the High-Cycle Fatigue strength of different load cases, short or long patches are to be used. In addition, the results show that some potential patch designs are not optimal in any case, and therefore can be completely excluded. Finally, the benefits of the unified interface of the engineering optimization framework are emphasized. Different optimization settings of a patch repair optimization are presented and compared utilizing the hypervolume metric. Concluding remarks on the potential of computational methods for improved repair design and an outlook on future maintenance of blisks complete this work.