Machine design is a wide and inherently multidisciplinary area. The research at this division is directed towards formal methods for systems engineering and design, product development methods, industrial design and optimization design methods. The philosophy is to develop design methods which utilize state of the art modeling methods in order to design complex and multidisciplinary products. Click on the bellow buttons to further explore our research areas. Click here to view the connection between our research fields and our applied applications.
Multidisciplinary Design Optimization
In the design of complex and tightly integrated engineering products, it is essential to be able to handle interactions among different subsystems of multi-disciplinary nature. Typical examples of such products include transportation vehicles like automobiles, trains, aircraft, and mechatronic machines like industrial robots. Hence it is necessary to combine models from several disciplines in order to simulate and analyze the behavior of the entire system and understand characteristics of sub-systems in a system context. Furthermore, to achieve an optimal design, a product must be treated as a complete system instead of developing subsystems independently. Optimizing each subsystem by itself would most likely lead to a sub-optimized system. Multidisciplinary design optimization (MDO) has been established as a convincing concurrent design optimization technique in development of such complex products. MDO has found broad application in various complex products with promising results. This is especially true in the aerospace domain where MDO is a widespread methodology in both academy and industry.
It has been identified that one of the more challenging aspects of MDO is modeling and simulation. The division of Machine Design has concentrated its efforts in this particular field during the last decade, and has developed a generic methodology to allow Geometry Based Automation (GBA). By implementing this method, geometry models are automatically generated within the MDO process. The methodology permits the geometry model to be altered during optimization, and hence the automated geometry generation is a key enabler for so-called geometry-in-the-loop multidisciplinary design. Furthermore, GBA serves as a framework integrator for other engineering tools, as it enables programs such as Finite Element (FE) and Computational Fluid Dynamic (CFD) analyses and other simulation programs to refer to a common geometric model.
Metamodels (also known as surrogate models) are computationally efficient numerical models of system outputs or other model outputs. It is unrealistic to perform exhaustive analyses for time-demanding simulations due to limited available time and occupation of computer power. This is solved by creating approximative numerical models of the original models and then performing the analyses on the metamodels instead of the original models.
At the division of Machine Design, the research is focused on the evaluation and application of metamodels for evaluating design concepts at early stages of the design process. Metamodels such as kriging and polynomial response surfaces are used for robust design optimization, reliability based design optimization and to lower the required number of model simulations for optimization algorithms.
Optimization algorithms are widely used by the industry for designing new products. By performing optimizations, the user get suggestions of optimal design parameters from the computer. Even though the engineering intuition and experience may be sufficient to suggest optimal designs, the optimization may present hitherto unthought solutions.
The division of Machine Design focuses on developing and evaluating different algorithms for performing optimizations of industrial design problems. Endeavors have foremost been made to modify the Nealder Mead Simplex algorithm to increase its robustment and increase its applications for problems where the objective function varies with time. Research is also performed to enhance its performance for computationally demanding simulations by using metamodels.
When developing new products, there are always uncertainties present. They may include operational conditions, manufacture tolerances, measurements errors and model errors. Consequently, design decisions need to be taken even though uncertainties are present, and it is therefore desirable to estimate how the uncertainties affects the different concepts. Uncertainty analysis strives at estimating how uncertain the values of the outputs of a model are when uncertainties in the model and/or its inputs are present.
The research at the division of Machine Design aims at evaluating and developing different methods for handling uncertainties, enabling the use of uncertainty analysises in earlier stages of the design process. This includes methods for propagating the uncertainty from the uncertainty sources to the model output, such as Monte Carlo simulations and Latin Hypercube Sampling.
Last updated: 2011-10-26