Information-Theoretic Methods for Modularity in Engineering Design

Mechanical Bus for Modular Product Design

PurposeWhile the beneficial impact of supply chain integration (SCI) and modular product design are generally acknowledged, few empirical studies have examined how an organization can achieve better performance through SCI with modular product design. The purpose of this paper is to examine the relationship between SCI and modular product design, as well as their impact on product performance.Design/methodology/approachBy surveying 251 manufacturers in Hong Kong, structural equation modelling is used to test the research constructs and the hypothesized model.FindingsThe results confirm that information sharing, product co‐development and organizational coordination are crucial organizational processes within SCI. Companies that have high levels of product modularity appear to be good at product co‐development and organizational coordination directly and at information sharing indirectly. Furthermore, companies that have high levels of product co‐development or product modularity appear to have better product performance.Research limitations/implicationsThis paper theoretically and empirically identifies three specific organizational processes within SCI (information sharing, product co‐development and organizational coordination), which affect modular product design and product performance. These more specific findings were previously absent from the literature. However, the study is limited to the cross‐sectional nature of a survey study, the operationalization of SCI and product modularity, and the nature of the product types.Originality/valueThis paper empirically examines the relationships between SCI and product modularity, which has seldom been attempted in previous research. It clearly identifies exactly which processes within SCI are directly and indirectly related to product modularity.

<p>Classification aims to identify a class label of an instance according to the information from its characteristics or features. Unfortunately, many classification problems have a large feature set containing irrelevant and redundant features, which reduce the classification performance. In order to address the above problem, feature selection is proposed to select a small subset of relevant features. There are three main types of feature selection methods, i.e. wrapper, embedded and filter approaches. Wrappers use a classification algorithm to evaluate candidate feature subsets. In embedded approaches, the selection process is embedded in the training process of a classification algorithm. Different from the other two approaches, filters do not involve any classification algorithm during the selection process. Feature selection is an important process but it is not an easy task due to its large search space and complex feature interactions. Because of the potential global search ability, Evolutionary Computation (EC), especially Particle Swarm Optimization (PSO), has been widely and successfully applied to feature selection. However, there is potential to improve the effectiveness and efficiency of EC-based feature selection. The overall goal of this thesis is to investigate and improve the capability of EC for feature selection to select small feature subsets while maintaining or even improving the classification performance compared to using all features. Different aspects of feature selection are considered in this thesis such as the number of objectives (single-objective/multi-objective), the fitness function (filter/wrapper), and the searching mechanism. This thesis introduces a new fitness function based on mutual information which is calculated by an estimation approach instead of the traditional counting approach. Results show that the estimation approach works well on both continuous and discrete data. More importantly, mutual information calculated by the estimation approach can capture feature interactions better than the traditional counting approach. This thesis develops a novel binary PSO algorithm, which is the first work to redefine some core concepts of PSO such as velocity and momentum to suit the characteristics of binary search spaces. Experimental results show that the proposed binary PSO algorithm evolve better solutions than other binary EC algorithms when the search spaces are large and complex. Specifically, on feature selection, the proposed binary PSO algorithm can select smaller feature subsets with similar or better classification accuracies, especially when there are a large number of features. This thesis proposes surrogate models for wrapper-based feature selection. The surrogate models use surrogate training sets which are subsets of informative instances selected from the training set. Experimental results show that the proposed surrogate models assist PSO to reduce the computational cost while maintaining or even improving the classification performance compared to using only the original training set. The thesis develops the first wrapper-based multi-objective feature selection algorithm using MOEA/D. A new decomposition strategy using multiple reference points for MOEA/D is designed, which can deal with different characteristics of multi-objective feature selection such as highly discontinuous Pareto fronts and complex relationships between objectives. The experimental results show that the proposed algorithm can evolve more diverse non-dominated sets than other multi-objective algorithms. This thesis introduces the first PSO-based feature selection algorithm for transfer learning. In the proposed algorithm, the fitness function uses classification performance to reduce the differences between domains while maintaining the discriminative ability on the target domain. The experimental results show that the proposed algorithm can select feature subsets which achieve better classification performance than four state-of-the-art feature-based transfer learning algorithms.</p>

Researchers have expanded the definition of product modularity from function-based modularity to life-cycle process-based modularity, and developed a measure of product modularity and validated a corresponding modular product design method. However, a correct modularity measure and modular design method are not enough to realize modular product design. How the measure and design method are used, especially the role of product representation, is an important aspect of modular design and imperative for realizing the promised cost savings of modularity. In this paper, we develop a representation that includes similarity and dependency for assembly modularity. The similarity representation is based on assembly cost factors including tool changes and fixture changes. The process-based design dependency representation is based on design features that impact the assembly process (faces, fasteners, assembly interference, etc.). This representation captures the cost benefits of modularity and can be extended to other life-cycle processes. A set of products is used to show the application of this representation in association with a modularity measure and modular design method.

Both modular product design and reconfigurable manufacturing have a great potential to enhance responsiveness to market changes and to reduce production cost. However, the two issues have thus far mostly been investigated separately, thereby causing possible mismatch between the modular product structure and the manufacturing or assembly system. Therefore, the potential benefits of product modularity may not be materialized due to such mismatch. For this reason, this paper presents a concurrent approach to the product module selection and assembly line design problems to provide a set of harmonic solutions to the two problems and hence avoid the mismatch between design and manufacturing. The integrated nature of the problem leads to several noncommensurable and often conflicting objectives. The modified Chebyshev goal programming approach is applied to solve the multi-objective problem. A genetic algorithm is further developed to provide quick and near-optimum solutions. The proposed approach and the solution procedure have been applied to an ABS motor problem. The performance of the genetic algorithm has also been examined.

This paper presents a Wavelet-based Artificial Neural Network (WANN) approach to model rainfall-runoff process of the Delaney Creek and Payne Creek watersheds with distinct hydro-geomorphological characteristics, located in Florida. Wavelet is utilized to handle the multi-frequency characteristics of the process in daily and monthly time scales. Thus, rainfall and runoff time series were decomposed into several sub-series by various mother wavelets. Due to multiple components obtained through wavelet decomposition, input sets to the Feed Forward Neural Network (FFNN) were enhanced. The application of two information content based criteria (i.e., entropy, H, and mutual information, MI) to select more reliable input sets (among all potential input sets) and to have better insight into the physics of process is considered as the basic innovation of the study which led to a more accurate and compact model. The increase in the number of input of the FFNN might lead to a complex structure and low performance. The results demonstrated that MIÂ Â as a supervised feature extraction criterion could lead to more reliable outcomes due to its non-linear nature. Furthermore, results indicated the superiority of proposed entropy-based WANN model (EWANN) in comparison to simple FFNN. Moreover, multi-step-ahead FFNN, conventional WANN and classic Auto Regressive Integrated Moving Average with eXogenous inputs (ARIMAX) models could not reveal appropriate forecasting results with regard to EWANN model. The superiority of the EWANN over the WANN and FFNN models is not only in terms of efficiency criteria, but also due to its appropriate ability to provide information about the physics of the process. The consequences of EWANN for rainfall-runoff modeling of two watersheds revealed that the proposed EWANN could simulate the process of a small and flat sub-basin slightly reliable than a sloppy and wild watershed. The poor outcome of monthly modeling in regard to daily modeling might be due to involvement of more uncertainty in the monthly data.

Researchers have expanded the definition of product modularity from function-based modularity to life-cycle process-based modularity. In parallel, measures of product modularity have been developed as well as corresponding modular product design methods. However, a correct modularity measure and modular design method are not enough to realize modular product design. To apply the measure and design method correctly, product representation becomes an important aspect of modular design and imperative for realizing the promised cost savings of modularity. In this paper, a representation for retirement process-based modular design has been developed. Built upon previous representations for assembly and manufacturing-based product design, the representation includes a process similarity matrix and a process dependency matrix. The retirement process-based similarity is based on the similarity in components’ post-life intents (recycling, reuse, disposal), and either the degree of their material compatibility if the components will be recycled, or their disassembly direction or disassembly tools if they need to be disassembled from each other for retirement. Process similarity within a module leads to increased process efficiency (the elimination of non-value added tasks) from the sharing of tooling/equipment. Retirement process-based dependency is developed based on disassembly difficulty, one aspect of the physical interactions between components. Retiring components together as a module to eliminate disassembly and differential processing and reducing the disassembly difficulty between the modules can increase the efficiency of the retirement process. We have first presented which process elements we should consider for defining retirement process similarity and dependency, and then constructed the respective similarity and dependency factors tables. These tables include similarity and dependency factors, which, along with their quantifications, are used to determine a product’s modular architecture to facilitate the retirement process. Finally, a fishing reel is used to illustrate how to apply these factors tables to generate the similarity and dependency matrices that represent a product for retirement-process based modular design. Using these representations as input to the DSM-based modular design methods, we can achieve a design with a modular architecture that improves the retirement process efficiency and reduces retirement costs.

Location‐only and use‐availability data: analysis methods converge

Transitions towards a circular economy (CE) are based on the complementary development of novel circular business models and innovative product designs. Among the most prominent design strategies are modular product architectures that break products down into components that can be independently and easily replaced or reused. So far, most research has focused on technological and business model requirements for modular products but neglected the role of consumers. With the example of smartphones, the chapter shows how the co-shaping of technological and consumer trends does not provide an easy starting point for the modularity of smartphones in a CE. It investigates modularity as a sociotechnical arrangement embedded into people’s everyday lives and reveals that different forms of modular designs and accompanying services need to be implemented to move towards a more sustainable use of smartphones. The chapter concludes that an efficient and sufficient use of modularity as a part of transitions towards a CE cannot be enabled by modular product designs and services alone but would need more political regulation and standardisation.

: Calculating the amount of information about a random function contained in another random function has important uses in communication theory. An expression for the mutual information for continuous time random processes has been given by Gelfand and Yaglom, Chiang, and Perez by generalizing Shannon's result in a natural way. Under a condition of absolute continuity of measures the continuous time expression has the same form as Shannon's result. For two Gaussian processes Gelfand and Yaglom express the mutual information in terms of a mean square estimation error. We generalize this result to diffusion processes and express the solution in a different form which is more naturally related to a corresponding filtering problem. We also use these results to calculate some information rates.

Modular product design, as part of a product family strategy, is a key methodology for mass customisation and significantly impacts production line configuration. Common modules shared across product families, generations, and the entire product platform have varying lifecycles due to constant changes in market demand. Therefore, it is crucial to design a production line that can quickly respond to product change requirements while maintaining flexibility in both product and process variants. The objective of this research is to develop a hybrid decision-making framework for determining an optimal production line configuration based on product modularity and lifecycle. In the proposed framework, three scenarios related to production line configuration are modelled as batch, mixed-model, and intermittent production lines based on specific product lifecycle. After modelling the three lines, data on processing times and inventory levels are gathered, while assumptions for additional data are made with industry collaborators. The proposed models are used as evaluation criteria in a Multi-Criteria Decision-Making (MCDM) framework to determine an optimal production line configuration by integrating product modules and lifecycle. The proposed research demonstrates the performance of the three scenarios quantitatively with eight sub-performance indicators, utilising a hybrid MCDM and real production line simulation data.

Design of a dynamic model of genes with multiple autonomous regulatory modules by evolutionary computations.

In evolutionary robotics, an encoding of the control software that maps sensor data (input) to motor control values (output) is shaped by stochastic optimization methods to complete a predefined task. This approach is assumed to be beneficial compared to standard methods of controller design in those cases where no a priori model is available that could help to optimize performance. For robots that have to operate in unpredictable environments as well, an evolutionary robotics approach is favorable. We present here a simple-to-implement, but hard-to-pass benchmark to allow for quantifying the “evolvability” of such evolving robot control software towards increasing behavioral complexity. We demonstrate that such a model-free approach is not a free lunch, as already simple tasks can be unsolvable barriers for fully open-ended uninformed evolutionary computation techniques. We propose the “Wankelmut” task as an objective for an evolutionary approach that starts from scratch without pre-shaped controller software or any other informed approach that would force the behavior to be evolved in a desired way. Our main claim is that “Wankelmut” represents the simplest set of problems that makes plain-vanilla evolutionary computation fail. We demonstrate this by a series of simple standard evolutionary approaches using different fitness functions and standard artificial neural networks, as well as continuous-time recurrent neural networks. All our tested approaches failed. From our observations, we conclude that other evolutionary approaches will also fail if they do not per se favor or enforce the modularity of the evolved structures and if they do not freeze or protect already evolved functionalities from being destroyed again in the later evolutionary process. However, such a protection would require a priori knowledge of the solution of the task and contradict the “no a priori model” approach that is often claimed in evolutionary computation. Thus, we propose a hard-to-pass benchmark in order to make a strong statement for self-complexifying and generative approaches in evolutionary computation in general and in evolutionary robotics specifically. We anticipate that defining such a benchmark by seeking the simplest task that causes the evolutionary process to fail can be a valuable benchmark for promoting future development in the fields of artificial intelligence, evolutionary robotics, and artificial life.

Evolutionary computation (EC) is widely applied to various kinds of combinatorial optimization problems. EC is generally a time-consuming paradigm because it needs much trial and error. To accelerate ECs, some modification methods of the genetic operator have been proposed such as improving mutation and recombination of chromosomes and/or their control parameters, and so on. Through these modifications, ECs can find the suboptimal solutions in the early generations. In spite of these improvements, ECs still require much time to obtain the solution in many practical engineering applications. Fitness evaluation usually spends most of the computational time. This paper presents a new approach for the acceleration of ECs by reducing the time for fitness evaluation. Saving the time for fitness evaluation results in accelerating the ECs in the time domain. In the proposed method, only one individual of the population is actually evaluated in each generation. Fitness values for the rest of the population are estimated with simple calculation. Although the errors of estimation may decelerate the ECs in the generation domain, the time saving in the evaluation scheme exceeds the deceleration. As a result, we can obtain suboptimal solutions relatively faster. The simulation results of the designed fuzzy logic controller using GA show the effectiveness of the proposed method to accelerate evolution in the time domain. The simulation results applied to De Jong's (1975) Test Function show the applicability to various problems of the proposed method.

The design of evolutionary computations based on schema processing, minimizing expected losses, and emphasizing certain genetic operators has failed to provide robust optimization performance. Recently, fitness distribution analysis has been proposed as an alternative tool for exploring operator behavior and designing efficient evolutionary computations. For example, the step size of a single parent variation operator, such as the Gaussian mutation operator, determines the corresponding probability of finding better solutions and the expected improvement that will be obtained. The paper analyses the utility of Gaussian, Cauchy, and mean mutation operators when a parent is located near a local extrema of a continuous objective function that is to be optimized.