The traditional production paradigm of large batch production does not offer flexibility toward satisfying the requirements of individual customers. A new generation of smart factories is expected to support new multivariety and small-batch customized production modes. For this, artificial intelligence (AI) is enabling higher value-added manufacturing by accelerating the integration of manufacturing and information communication technologies, including computing, communication, and control. The characteristics of a customized smart factory are: self-perception, operations optimization, dynamic reconfiguration, and intelligent decision-making. The AI technologies will allow manufacturing systems to perceive the environment, adapt to the external needs, and extract the process knowledge, including business models, such as intelligent production, networked collaboration, and extended service models. This article focuses on the implementation of AI in customized manufacturing (CM). The architecture of an AI-driven customized smart factory is presented. Details of intelligent manufacturing devices, intelligent information interaction, and construction of a flexible manufacturing line are showcased. The state-of-the-art AI technologies of potential use in CM, that is, machine learning, multiagent systems, Internet of Things, big data, and cloud-edge computing, are surveyed. The AI-enabled technologies in a customized smart factory are validated with a case study of customized packaging. The experimental results have demonstrated that the AI-assisted CM offers the possibility of higher production flexibility and efficiency. Challenges and solutions related to AI in CM are also discussed.

Rainfall affects local water quantity and quality. A data-mining approach is applied to predict rainfall in a watershed basin at Oxford, Iowa, based on radar reflectivity and tipping-bucket (TB) data. Five data-mining algorithms, neural network, random forest, classification and regression tree, support vector machine, andk-nearest neighbor, are employed to build prediction models. The algorithm offering the highest accuracy is selected for further study. Model I is the baseline model constructed from radar data covering Oxford. Model II predicts rainfall from radar and TB data collected at Oxford. Model III is constructed from the radar and TB data collected at South Amana (16 km west of Oxford) and Iowa City (25 km east of Oxford). The computation results indicate that the three models offer similar accuracy when predicting rainfall at current time. Model II performs better than the other two models when predicting rainfall at future time horizons.

This paper presents a dynamic predictive-optimization framework of a nonlinear temporal process. Data-mining (DM) and evolutionary strategy algorithms are integrated in the framework for solving the optimization model. DM algorithms learn dynamic equations from the process data. An evolutionary strategy algorithm is then applied to solve the optimization problem guided by the knowledge extracted by the DM algorithm. The concept presented in this paper is illustrated with the data from a power plant, where the goal is to maximize the boiler efficiency and minimize the limestone consumption. This multiobjective optimization problem can be either transformed into a single-objective optimization problem through preference aggregation approaches or into a Pareto-optimal optimization problem. The computational results have shown the effectiveness of the proposed optimization framework.

A dynamic predictive-control model of a nonlinear and temporal process is considered. Evolutionary computation and data mining algorithms are integrated for solving the model. Data-mining algorithms learn dynamic equations from process data. Evolutionary algorithms are then applied to solve the optimization problem guided by the knowledge extracted by data-mining algorithms. Several properties of the optimization model are shown in detail, in particular, a selection of regressors, time delays, prediction and control horizons, and weights. The concepts proposed in this paper are illustrated with an industrial case study in combustion process.

The assemble-to-order (ATO) production strategy considers a tradeoff between the size of a product portfolio and the assembly lead time. The concept of modular design is often used in support of the ATO strategy. Modular design impacts the assembly of products and the supply chain, in particular, the storage, transport, and production are affected by the selected modular structure. The demand for products in a product family impacts the cost of the supply chain. Based on the demand patterns, a mix of modules and their stock are determined by solving an integer programming model. This model cannot be optimally solved due to its high computational complexity and, therefore, two heuristic algorithms are proposed. A simulated annealing algorithm improves on the previously generated solutions. The computational results reported in this paper show that significant savings could be realized by optimizing the composition of modules. The best performance is obtained by a simulated annealing combined with a heuristic approach.

In this paper, a data-mining approach is used to develop a model for optimizing the efficiency of an electric-utility boiler subject to operating constraints. Selection of process variables to optimize combustion efficiency is discussed. The selected variables are critical for control of combustion efficiency of a coal-fired boiler in the presence of operating constraints. Two schemes of generating control settings and updating control variables are evaluated. One scheme is based on the controllable and noncontrollable variables. The second one incorporates response variables into the clustering process. The process control scheme based on the response variables produces the smallest variance of the target variable due to reduced coupling among the process variables. An industrial case study and its implementation illustrate the control approach developed in this paper.

Product design engineering is undergoing a transformation from informal and largely experience-based discipline to a science-based domain. Computational intelligence (CI) offers models and algorithms that can contribute greatly to design formalization and automation. This paper surveys CI concepts and approaches applicable to product design engineering. Taxonomy of the surveyed literature is presented according to the generally recognized areas in both product design engineering and CI. Some research issues that arise from the broad perspective presented in the paper have been signaled but not fully pursued. No survey of such a broad field can be complete; however, the material presented in the paper is a summary of state-of-the-art CI concepts and approaches in product design engineering.

Manufacturing companies are focusing on mass customization. Delivering products that meet the requirements of individual customers complicates the production process, and diminishes the benefits of the economy of scale. By exploring commonality among products, this complexity can be significantly reduced. To determine product configurations sought by the customers and to produce them in large quantities, a new approach is proposed. The proposed approach uses a modified k-means clustering algorithm to analyze past sales data for capturing prime product configurations. The most suitable configurations are selected by solving an integer-programming model or using a sorting-based algorithm. The proposed approach was tested with an industrial case study involving sales data of large trucks collected over a period of one year.

The term innovation resonates broadly in the cyberspace, books, and journals. A careful analysis of the literature indicates that the knowledge of the underlying science of innovation is limited. Innovation in any domain can be enhanced by principles and insights from different disciplines. However, the process of identifying the linkages between the diverse disciplines and the target domain is not well understood. The innovation process and conditions triggering innovation set the stage for economic progress. The ideas outlined in the paper provide a roadmap for future discovery of innovation science.

In this paper, a data-mining approach is applied to optimize combustion efficiency of a coal-fired boiler. The combustion process is complex, nonlinear, and nonstationary. A virtual testing procedure is developed to validate the results produced by the optimization methods. The developed procedure quantifies improvements in the combustion efficiency without performing live testing, which is expensive and time consuming. The ideas introduced in this paper are illustrated with an industrial case study.

Analytical knowledge is distributed among domain experts, analysts, and data-storage systems. Extracting such knowledge from databases is of interest to corporations. The traditional top-down development of corporate memory is not appropriate for modern organizations because of the distributed nature of information. This paper proposes models of analytical knowledge and new ways of developing corporate memory by using an extensible markup language (XML). It aims at efficient exploration of useful knowledge by mining the Web. The proposed approach of modeling analytical knowledge is explicit and sharable. The concepts introduced in the paper have been demonstrated with a manufacturing case study.

Clinical medicine is facing a challenge of knowledge discovery from the growing volume of data. In this paper, a data mining algorithm (G-algorithm) is proposed for extraction of robust rules that can be used in clinical practice for better understanding and prevention of unwanted medical events. The G-algorithm is applied to the data set obtained for children born with a malformation of the heart (univentricular heart). As the result of the Fontan surgical procedure, designed to palliate the children, 10%-35% of patients postoperatively develop an arrhythmia known as the intra-atrial reentrant tachycardia. There is an obvious need to identify the children that may develop the tachycardia before the surgery is performed. Prior attempts to identify such children with statistical techniques have been unrewarding. The G-algorithm discussed in this paper shows that there exists an unambiguous relationship between measurable features and the tachycardia. The data set used in this study shows that, for 78.08% of infants, the occurrence of tachycardia can be accurately predicted. The authors' prior computational experience with diverse medical data sets indicates that the percentage of accurate predictions may become even higher if data on additional features is collected for a larger data set.

This paper discusses an industrial case study in which data mining has been applied to solve a quality engineering problem in electronics assembly. During the assembly process, solder balls occur underneath some components of printed circuit boards. The goal is to identify the cause of solder defects in a circuit board using a data mining approach. Statistical process control and design of experiment approaches did not provide conclusive results. The paper discusses features considered in the study, data collected, and the data mining solution approach to identify causes of quality faults in an industrial application.

In the paper, a prediction problem encountered in engineering design is considered. The problem is solved with a data mining approach. The first of the two algorithms presented in the paper generates a solution, while the second algorithm validates this solution. Only confirmed solutions are accepted.

The researchers and practitioners of today create models, algorithms, functions, and other constructs defined in abstract spaces. The research of the future will likely be data driven. Symbolic and numeric data that are becoming available in large volumes will define the need for new data analysis techniques and tools. Data mining is an emerging area of computational intelligence that offers new theories, techniques, and tools for analysis of large data sets. In this paper, a novel approach for autonomous decision-making is developed based on the rough set theory of data mining. The approach has been tested on a medical data set for patients with lung abnormalities referred to as solitary pulmonary nodules (SPNs). The two independent algorithms developed in this paper either generate an accurate diagnosis or make no decision. The methodolgy discussed in the paper depart from the developments in data mining as well as current medical literature, thus creating a variable approach for autonomous decision-making.

Modularity refers to the use of common units to create product variants. This paper aims at the development of models and solution approaches to the modularity problem for mechanical, electrical, and mixed process products (e.g., electromechanical products). To interpret various types of modularity, e.g., component-swapping, component-sharing and bus modularity, a matrix representation of the modularity problem is presented. The decomposition approach is used to determine modules for different products. The representation and solution approaches presented are illustrated with numerous examples. The paper presents a formal approach to modularity allowing for optimal forming of modules even in the situation of insufficient availability of information. The modules determined may be shared across different products.

To respond to the challenge of agile manufacturing, companies are striving to provide a large variety of products at a low cost. Product modularity allows production of different products by combining standard components. One of the characteristics of modular products is that they share the same assembly operations for a part of their structure. The special structure of modular products provides challenges and opportunities for the design of assembly systems. Given a family of modular products, designing low cost assembly systems is an important problem. In this paper, an approach for the design of assembly systems for modular products is proposed. The assembly system is decomposed into two subsystems based on the structure of modular products. The configuration problem of the assembly system is formulated and solved by a tabu search based algorithm.

In this paper, an analytical approach for the availability evaluation of cellular manufacturing systems is presented, where a manufacturing system is considered operational as long as its production capacity requirements are satisfied. The advantage of the approach is that constructing a system level Markov chain (a complex task) is not required. A manufacturing system is decomposed into two subsystems, i.e. machining system and material handling system. The machining subsystem is in turn decomposed into machine cells. For each machine cell and material handling subsystem, a Markovian model is derived and solved to find the probability of a subset of working machines in each cell, and a subset of the operating material handling carriers that satisfies the manufacturing capacity requirements. The overall manufacturing system availability is obtained using a procedure presented in the paper. The novelty of the approach is that it incorporates imperfect coverage and imperfect repair factors in the Markovian models. The approach is used to evaluate transient and steady-state performance of three alternative designs based on an industrial example. Detailed discussion of the results and the impact of imperfect coverage and imperfect repair on the availability of the manufacturing system is presented. Possible extensions of the work and software tools available for model analysis are also discussed.

A design process can be represented as a network of design activities. A number of design projects may be undertaken simultaneously. This paper deals with the problem of scheduling design activities of multiple design projects competing for the limited available resources. The problem of determining a schedule subject to precedence and resource constraints is difficult to solve. It becomes even more complex when unforeseen changes are considered, for example, in the level of resources. Therefore, the scheduling problem is decomposed into a series of multidimensional (multiresource) knapsack problems. Due to high computational complexity of the multidimensional knapsack problem, two solution procedures are proposed.

The design process can be considered as a series of design phases. Each design phase can be further divided into number of design activities. In the network of design activities, the design activities generate information which is used by the succeeding activities. Since importance of the input information available from different activities changes with respect to the requirement of the design activities, a pull system approach for the management of design activities is proposed. Due to the concurrent consideration of many factors in the design process, it is required that the design activities are performed by teams of experts. In addition to the multiple resource requirement, the uncertainty in the weights and duration of design activities makes the problem of scheduling design activities complex. A heuristic algorithm is proposed to solve the problem considered in this paper.

This paper presents a risk assessment strategy for managing the design process. The proposed method intends to provide the foundation for an intelligent system that can assist design engineers and managers of the design process. The IDEF3 methodology is used to describe the concurrent design process. The logical semantics of IDEF3 provide a much more detailed description than traditional project management methods, such as CPM. The IDEF3 model is then used to identify alternative scenarios that result in the successful completion of a design project. Using estimates of the likelihood that a scenario will fail to produce a successful design and the consequences that result, alternative scenarios are evaluated. A working example is provided, and the architecture of an intelligent system for risk assessment is discussed.

Problems in engineering design are constraint-oriented and often involve multiple perspectives. Designers must consider not only functional requirements of a product but also its life-cycle perspectives. Each perspective has its own set of constraints which may contain conflicting or unsatisfied requirements. A human designer can not be aware of all constraints and design alternatives all the time. The objective of this paper is to develop a methodology to assist designers in negotiation of constraints. A network model is proposed to represent relationships among design variables. An algorithm is developed to derive dependencies between design variables and goals. Based on the dependencies obtained, design modifications are generated for resolving conflicts. In the second part of the paper, a fuzzy-logic-based approach is used to model imprecise dependencies between variables in the case when no sufficient quantitative information is available. The approach proposed can be used to increase the amount of information provided to the designers for making decisions.>

The incorporation of alternative operations in a scheduling system increases the utilization rate of resources and reduces the makespan of manufacturing products. A heuristic algorithm is developed for a scheduling problem with and without alternative operations. The effect of alternative operations on the performance of schedules generated are studied with five dispatching rules. The testing effort involves 240 scheduling problems obtained for randomly generated data. The computational results show that the most dissimilar resources (MDR) dispatching rule for the case with alternative operations performs best among the dispatching rules tested. The quality of schedules (makespan, utilization rate of resources) generated with any dispatching rule improves when alternative operations are used.>

Process planning for a part involves preparation of a plan that outlines the processing route, operations, machine tools, and so on. The decomposition of the process-planning task into eight phases is discussed. For each phase, problems and suitable solution approaches are discussed. Most process-planning problems are solved using a knowledge-based approach or a combination of optimization and knowledge-based approaches. The knowledge-based approach solves the qualitative subproblems, and the optimization perspective tackles the quantitative aspect of the subproblems. The concepts presented are illustrated with examples.>

An aggregate scheduling problem for a flexible manufacturing system which consists of a machining and an assembly subsystem is presented. The aggregate schedule developed is used as an input to a real-time scheduling problem discussed previously by the author (Proc. 2nd ORSA/TIMS Conf. on Flexible Manuf. Syst: Oper. Res. Methods and Appl., Elsevier NY, 1986, pp.521-32). Although features of the flexible manufacturing environment are not explicit at the aggregate scheduling level, they are incorporated at the real-time scheduling level. The single-product and N-product scheduling problems are considered. To solve the problems two optimal scheduling algorithms are developed. Each algorithm is illustrated with a numerical example. The results are also applicable to the single-batch and N-batch scheduling problems.>

A problem-solving approach involving the integration of expert systems and optimization techniques is presented. A class of expert systems called tandem expert systems is introduced. Three variants of the tandem expert system, the data-reducing, model-based, and model-modifying expert systems, are explained with examples. It is emphasized that optimization techniques can be used more frequently in future expert systems.>

An approach to decomposition of manufacturing systems known as the group technology (GT) is surveyed. GT makes it possible to cluster machines into machine cells and parts into part families. There are two basic methods used for solving the GT problem: classification and modeling. Two variations of the classification method, visual and coding, are briefly discussed. The matrix and mathematical programming formulations of the GT problem are presented along with algorithms for solving them.>

Most of the planning models for automated manufacturing systems are based on the assumption that for each part there is only one process plan available. A more realistic point of view is taken that for each part a number of different process plans are generated, each of which may require specific types of tools and auxiliary devices such as fixtures, grippers, and feeders. A model for the selection of a set of process plans with the minimum corresponding manufacturing cost and minimal number of tools and auxiliary devices is formulated. The developed model for m parts is equivalent to the problem of finding the maximum clique in an m-partite graph with the minimum corresponding cost. Heuristic algorithms and numerical results are discussed.>

In this paper the existing approaches to design of expert scheduling systems are briefly described. Features of flexible manufacturing environment are explored. To take advantage of the progress in AI and optimization theory two new architectures: a single-model tandem architecture (Kusiak [5]) and multimodel tandem architecture of expert scheduling systems are discussed. Based on these architectures two expert systems for scheduling flexible manufacturing systems are designed.

Clustering of large-scale binary matrices requires a considerable computational effort. In some cases this effort is lost since the matrix is not decomposable into mutually separable submatrices. A cluster identification algorithm which has relatively low computational time complexity O(2mn) is developed. It allows checking for the existence of clusters and determines the number of mutually separable clusters. A modified cluster identification algorithm for clustering nondiagonally structured matrices is also presented. The two algorithms are illustrated in numerical examples.

In this paper a structure of the manufacturing expert system is presented. This system is to control three subsystems: process planning, programming of robots and machines and production planning. One of the basic elements of the production planning system is the production scheduling. A scheduling framework together with the heuristic scheduling algorithm are developed. The algorithm is illustrated on a numerical example.

Most of the planning models for FMSs are based on the assumption that for each part there is only one process plan available. In this paper, we take the more realistic standpoint that for each part a number of different process plans could be used, each of which requires specific types of tools and fixtures. A model for the selection of a set of process plans with the minimum corresponding manufacturing cost and minimal number of tools and fixtures is formulated. Heuristic algorithms and numerical results are discussed.