Yanyan Wu

Prediction of Assembly Variation During Early Design

This paper presents the methods to move assembly variation analysis into early stages of aircraft development where critical partitioning, sourcing, and production decisions are often made for component parts that have not yet been designed. Our goal is to identify and develop variation prediction methods that can precede detailed geometric design and make estimates accurate enough to uncover major assembly risks. With this information in hand, design and/or manufacturing modifications can be made prior to major supplier and production commitments. In addition to estimation of the overall variation, the most significant contributors to assembly variation are also identified. In this paper, a generic framework for prediction of assembly variation has been developed. An efficient, top-down approach has been adopted. Instead of taking measurement everywhere, the variation analysis starts with airplane level requirements (e.g., load capabilities and orientation of horizontal/vertical stabilizers), and then assembly requirements (mainly geometric dimensioning and tolerancing callouts, quantifiable in quality control) are derived. Next the contributors to a particular assembly requirement are identified through data flow chain analysis. Finally, the major contributors are further characterized through a sensitivity study of metamodels or 3D variation analysis models. A case study of a vertical fin has been used to demonstrate the validity of the proposed framework. Multiple prediction methods have been studied and their applicability to variation analysis discussed. Simplified design simulation method and metamodel methods have been tested and the results are reported. Comparisons between methods have been made to demonstrate the flexibility of the analysis framework, as well as the utility of the prediction methods. The results of a demonstration test case study for vertical fin design were encouraging with modeling methods coming within 15% of deviation compared with the detailed design simulation.

Multimodal Industrial Inspection and Analysis

Diagnosis of complex engineering systems requires the use of multiple sensor sources to acquire information. In this paper we present a survey of multimodal data acquisition systems for nondestructive testing (NDT) and engineering analysis. We begin with a summary of the relative strengths and weaknesses of individual NDT modalities. Thereafter we present existing multimodal inspection hardware systems that use complementary NDT sensors. The advantages of such multimodal data acquisition over conventional single modality sensors in inspection and analysis are highlighted. Possible approaches to fuse complementary multimodal sensor data are discussed. We conclude with possible directions for the future development of multimodal inspection systems.

Methods to Improve the Accuracy of Registration for Multimodal Inspection of Mechanical Parts

Registration refers to the process of aligning corresponding features in images or point data sets in the same coordinate system. Multimodal inspection is a growing trend wherein an accurate measurement of the part is made by fusing data from different modalities. Registration is a key task in multimodal data fusion. The main problem with high-accuracy registration comes from noise inherent in the measurement data and the lack of the one-to-one correspondence in the data from different modalities. We present methods to deal with outliers and noise in the measurement data to improve registration accuracy. The proposed algorithms operate on point sets. Our method distinguishes between noise and accurate measurements using a new metric based on the intrinsic geometric characteristics of the point set, including distance, surface normal and curvature. Our method is unique in that it does not require a-priori knowledge of the noise in the measurement data, therefore fully automatic registration is enabled. The proposed methods can be incorporated into any point-based registration method. It was tested with the traditional ICP (Iterative Closest Point) algorithm with application to the data registration among point, image, and mesh data. The proposed method can be applied to both rigid and non-rigid registration.Copyright