|dc.description.abstract||Composite parts have been increasingly used in aerospace industry because of their high strength-to-weight ratio and stiffness-to-weight ratio. Due to the diversity of suppliers and variation in the fabrication process of composite parts, dimensional variability of composite fuselages inevitably exists. One of the critical challenges to reduce the dimensional variability of the assembly process is the complex property of composite materials. The traditional physical models applied to metal materials cannot be directly applied here. It is of high importance to develop systematic methodologies to conduct dimensional analysis, variation reduction, and optimal shape adjustment for the composite fuselages’ assembly process. Based on these motivations, this dissertation focuses on developing systematic methodologies for effective system modeling, quality control and variability reduction in the composite fuselages assembly process. These advanced methodologies enable a better understanding of the composite fuselage structure with actuator applied, a more accurate handling of the fuselage shape control system, and a more precise way to analyze the residual stress during the fuselage assembly process. This dissertation is organized as follows. Chapter 1 introduces a new shape control system with 10 actuators that is able to conduct dimensional shape adjustment before the airplane fuselage assembly process. In Chapter 2, a feasibility study is conducted to evaluate the proposed shape control concept. In the feasibility analysis, an accurate finite element model is developed to mimic the fabrication of composite fuselage, which includes the detailed materials setting, ply design, geometry and fixture structures. The finite element model is validated and calibrated based on physical experimental data with a real fuselage on the production floor. The results show that the single-plane with ten actuators scheme is feasible for shape control, and that actuators do not damage the fuselage. In Chapter 3, a surrogate model considering four types of uncertainties (actuator uncertainty, part uncertainty, modeling uncertainty, and unquantified uncertainty) has been developed for automatic optimal shape control of the system proposed in Chapter 2. A maximum likelihood estimation (MLE) algorithm is used for parameter estimation and response prediction. Afterwards, the surrogated model considering uncertainties is embedded into a feedforward control algorithm, which is achieved by conducting multivariate optimization to minimize the weighted summation of dimensional deviations of the response from the target. We show that the surrogate model considering uncertainties achieves satisfactory prediction performance and the automated optimal shape control system can significantly reduce the assembly cycle time with improved dimensional quality. In Chapter 4, two active learning algorithms are proposed for Gaussian process considering uncertainties, which are the variance-based weighted active learning algorithm and D-optimal weighted active learning algorithm. These active learning algorithms effectively reduce the number of samples needed to get accurate statistical models for industrial systems that have numerous uncertainties, such as input uncertainties, measurement errors, modeling uncertainties, and uncertainties from system parameters. The proposed algorithms investigate stochastic Kriging model and surrogate model considering uncertainties, with information measure of variance-based information and Fisher information. The algorithms have been applied to improving the predictive modeling for automatic shape control of composite fuselage. They can also be applied in other active learning scenarios for predictive models with multiple uncertainties. From Chapter 2 to Chapter 4, we have focused on optimal shape control system for single fuselage. After the two fuselages are adjusted to the target shape, they will be assembled together, and the actuator forces are released afterwards. The release of actuator forces results in the springback of two fuselages and the occurrence of residual stresses. In Chapter 5, we investigate the process of fuselages assembly via the proposed control system and develop an FEA platform to simulate the dimensional change and stress change during and after the assembly process. Instead of reversing the actuator forces, we use dynamic forces to simulate the springback of the fuselages after releasing the actuators. The proposed approach is more accurate compared to the traditional simulation approach. The case studies show that the assembly process with our new shape control system is applicable and the residual stresses are lower than the failure threshold. In Chapter 6, we discuss the potentials of future research. The automatic shape control system developed for single fuselage can be extended to variation reduction of multi-station fuselage assemblies. The surrogate modeling of residual stress is another area of interest, which could lead to fast stress prediction and optimization. Optimal fixture design is also of high importance due to the capability of stress reduction.
In summary, this thesis fuses the knowledge of statistics, mechanical engineering, and material science to obtain a better understanding and improvement of the composite fuselage assembly process. The methodologies and simulation platforms developed in this thesis have the potential to be applied to other advanced manufacturing systems.||