A GENERAL LOSS FUNCTION BASED OPTIMIZATION PROCEDURE FOR ROBUST DESIGN

1996 ◽  
Vol 25 (4) ◽  
pp. 255-276 ◽  
Author(s):  
BALAJI RAMAKRISHNAN ◽  
S. S. RAO
Keyword(s):  
Robust Design ◽  
Loss Function ◽  
Optimization Procedure ◽  
General Loss Function

Robust Design by Matching the Design With Manufacturing Variation Patterns

1994 ◽  
Author(s):  
Jyh-Cheng Yu ◽  
Kosuke Ishii
Keyword(s):  
Robust Optimization ◽  
Robust Design ◽  
Optimization Procedure ◽  
Theoretical Development ◽  
Future Application ◽  
Design Variables ◽  
Performance Variation ◽  
Heat Treated ◽  
Variation Patterns ◽  
Manufacturing Variation

Abstract This paper deals with robust design problems in which variations on design variables have significant correlation. Manufacturing errors often affect design variables with characteristic patterns, that is, the variations are coupled. Robust optimization seeks designs with optimal and robust performance. Designers should match the design to the Manufacturing Variation Patterns (MVP) in the constrained robust optimization procedure. This study focuses on matching the variation patterns found in typical manufacturing processes. It uses quadrature experimental design to approximate the performance variation within the patterns. We redefine the robust constraint activity for designs using MVP and propose our procedure to search for the robust feasible designs. Theoretical development of manufacturing variation matching leads to our case study of heat treated shaft design with minimum dimensional distortion. The paper also outlines our future application in injection molding gear design and challenge in the identification of nonlinear correlated MVP.

Robust Suspension System Design

1993 ◽  
Author(s):  
John E. Beard ◽  
John W. Sutherland
Keyword(s):  
System Design ◽  
Robust Design ◽  
Loss Function ◽  
Kinematic Model ◽  
Small Variation ◽  
Suspension System ◽  
Product Performance ◽  
Design Variables ◽  
Manufacturing Assembly ◽  
The Impact

Abstract Traditionally, levels for design variables are sought that produce optimal performance of a product. When manufacturing and assembly processes are used to realize the design intent, however, the product performance may differ from that envisioned during design. This is because the performance of a product is often very sensitive to manufacturing and assembly variations. This paper presents a methodology for robust design that incorporates the impact of manufacturing/assembly variations. The methodology characterizes the performance of a manufactured product via a loss function. The loss function measure is attractive from a robust design standpoint since it stresses both desirable performance on the average and small variation in performance from product to product. The design methodology is demonstrated through a suspension system design application. A model for the kinematic behavior of a suspension system is developed. The scrub rate is selected as the response of interest to demonstrate the methodology. The behavior of the kinematic model, in terms of the loss function, is approximated near a set point and levels of the design variables are sought that minimize the loss. An iterative procedure is described for optimizing the loss function. The application demonstrates that substantial improvements can be made in terms of actual manufactured product performance through the use of the methodology.

Development of a Robust Design Process Using a New Robustness Index

2005 ◽  
Author(s):  
K. H. Hwang ◽  
G. J. Park
Keyword(s):  
Design Process ◽  
Robust Design ◽  
Loss Function ◽  
The Other ◽  
The Third ◽  
Design And Manufacturing ◽  
Probability Of Success ◽  
Robustness Index ◽  
The Mean ◽  
Taguchi’S Loss Function

In product design and manufacturing, robust design leads to a product that has good quality. Robust design is reviewed in two categories: one is the process and the other is the robustness index. The process means efficient manipulation of the mean response and the variance. The robustness index indicates a measure of insensitiveness with respect to the variation. To improve existing methods, a three-step robust design (TRD) is proposed. The first step is “reduce the variance,” the second is “find multiple candidate designs,” and the third is “select the optimum robust design by using the robustness index,” Furthermore, a new robustness index is introduced in order to accommodate the characteristics of the probability of success in axiomatic design and the Taguchi’s loss function. The new robustness indices are compared with the existing ones. The developed robust design process is verified by examples and the results using the robustness index are compared with those of other indices.

Practical Robust Design Optimization Using Evolutionary Algorithms

2011 ◽  
Vol 133 (10) ◽  
Author(s):  
Amit Saha ◽  
Tapabrata Ray
Keyword(s):  
Evolutionary Algorithms ◽  
Robust Optimization ◽  
Design Optimization ◽  
Robust Design ◽  
Optimal Solution ◽  
Optimization Procedure ◽  
Computational Design ◽  
Robust Design Optimization ◽  
Function Evaluation ◽  
Engineering Design Problems

Robust design optimization (RDO) seeks to find optimal designs which are less sensitive to the uncontrollable variations that are often inherent to the design process. Studies using Evolutionary Algorithms (EAs) for RDO are not too many. In this work, we propose enhancements to an EA based robust optimization procedure with explicit function evaluation saving strategies. The proposed algorithm, IDEAR, takes into account a specified expected uncertainty in the design variables and then imposes the desired robustness criteria during the optimization process to converge to robust optimal solution(s). We pick up a number of Bi-objective engineering design problems from the standard literature and study them in the proposed robust optimization framework to demonstrate the enhanced performance. A cross-validation study is performed to analyze whether the solutions obtained are truly robust and also make some observations on how robust optimal solutions differ from the performance maximizing solutions in the design space. We perform a rigorous analysis of the key features of IDEAR to illustrate its functioning. The proposed function evaluation saving strategies are generic and their applications are worth exploring in other areas of computational design optimization.

scholarly journals Event-based backpropagation can compute exact gradients for spiking neural networks

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Timo C. Wunderlich ◽  
Christian Pehle
Keyword(s):  
Neural Networks ◽  
Loss Function ◽  
Spiking Neural Networks ◽  
Backpropagation Algorithm ◽  
General Loss Function ◽  
Gradient Based ◽  
Rigorous Study ◽  
Spiking Networks ◽  
Event Based ◽  
First Time

AbstractSpiking neural networks combine analog computation with event-based communication using discrete spikes. While the impressive advances of deep learning are enabled by training non-spiking artificial neural networks using the backpropagation algorithm, applying this algorithm to spiking networks was previously hindered by the existence of discrete spike events and discontinuities. For the first time, this work derives the backpropagation algorithm for a continuous-time spiking neural network and a general loss function by applying the adjoint method together with the proper partial derivative jumps, allowing for backpropagation through discrete spike events without approximations. This algorithm, EventProp, backpropagates errors at spike times in order to compute the exact gradient in an event-based, temporally and spatially sparse fashion. We use gradients computed via EventProp to train networks on the Yin-Yang and MNIST datasets using either a spike time or voltage based loss function and report competitive performance. Our work supports the rigorous study of gradient-based learning algorithms in spiking neural networks and provides insights toward their implementation in novel brain-inspired hardware.

scholarly journals FCN-Based 3D Reconstruction with Multi-Source Photometric Stereo

2020 ◽  
Vol 10 (8) ◽  
pp. 2914
Author(s):  
Ruixin Wang ◽  
Xin Wang ◽  
Di He ◽  
Lei Wang ◽  
Ke Xu
Keyword(s):  
3D Reconstruction ◽  
Loss Function ◽  
Classical Method ◽  
Solution Process ◽  
Photometric Stereo ◽  
Good Effect ◽  
Reconstruction Method ◽  
Convolutional Network ◽  
General Loss Function ◽  
3D Information

As a classical method widely used in 3D reconstruction tasks, the multi-source Photometric Stereo can obtain more accurate 3D reconstruction results compared with the basic Photometric Stereo, but its complex calibration and solution process reduces the efficiency of this algorithm. In this paper, we propose a multi-source Photometric Stereo 3D reconstruction method based on the fully convolutional network (FCN). We first represent the 3D shape of the object as a depth value corresponding to each pixel as the optimized object. After training in an end-to-end manner, our network can efficiently obtain 3D information on the object surface. In addition, we added two regularization constraints to the general loss function, which can effectively help the network to optimize. Under the same light source configuration, our method can obtain a higher accuracy than the classic multi-source Photometric Stereo. At the same time, our new loss function can help the deep learning method to get a more realistic 3D reconstruction result. We have also used our own real dataset to experimentally verify our method. The experimental results show that our method has a good effect on solving the main problems faced by the classical method.

Sign in / Sign up

Export Citation Format

Share Document