Cloud-Based Materials and Product Realization—Fostering ICME Via Industry 4.0

Mistree and co-authors [10] define design as the conversion of information that characterizes the needs and requirements for a product into knowledge about the product. The underlying philosophy in the definition of design by Mistree and co-authors [10] is that the designer starting with the functional requirements that is desired (the goal that designer wishes to achieve) should be able to work backwards to explore effective design solutions. This philosophy is adopted in this paper for design—as a goal-oriented activity. As noted by Gero [11], the designing involves transforming requirements—generally termed functions—into design descriptions. Decision-based design [10, 12] is a term coined to emphasize a different perspective to develop methods for design. The principal role of a human designer in decision-based design (DBD) is to make decisions given the information available. From an engineering perspective, decisions exclusively deal with allocation of resources in some form, usually as capital expenditures. Thus, the definition of a decision here is as “an irrevocable allocation of resources” [13]. We believe that there are two types of decisions that a human designer can make: selection and compromise decisions. A complex design can be represented by modeling a workflow of compromise and selection decisions, achieved using PDSIDES.

PDSIDES [14] is a “Knowledge-Based” Platform for Decision Support in the Design of Engineering Systems (PDSIDES) that is anchored in modeling decision-related knowledge with templates using ontologies to facilitate execution and reuse. The two primary constructs required for the realization of decisions within PDSIDES are [14] (1) decision support problem (DSP) construct and (2) ontology. Three types of platform users are defined according to the amount of knowledge they have for operating the decision template, namely, template creators, template editors, and template implementers. Template creators are domain experts and responsible for creating decision templates for original design, which requires the greatest novelty. Template editors are senior designers who have sufficient knowledge and experience in a specific domain and are responsible for editing (or tailoring) existing decision templates in adaptive design; this requires the original templates to be adapted for new applications. Template implementers are designers who have basic knowledge and typically little knowledge or interest in the analysis embodied in the template; they are responsible for executing existing decision templates that result in variant designs that require only parametric changes in the original decision templates. Ming and co-authors [15] present the ontologies developed for selection decision, compromise decision, and hierarchical coupled decisions using the software tool Protégé [16]. The classes and properties of the relevant ontologies are defined by Ming and co-authors (see [17, 18]). Ming and co-authors [14] present the very first version of platform PDSIDES which is web-based and deployed in the local server of the Systems Realization Laboratory at the University of Oklahoma.

PDSIDES has the potential to support a human designer with the identified core competencies for product realization when integrated with the cloud. In Fig. 1, we illustrate the concepts underlying the foundations and principles of Cloud-Based Platform for Decision Support in the Design of Engineering Systems (CB-PDSIDES) as proposed in this paper. To integrate PDSIDES with cloud and bring-in the concepts of cloud computing and collaboration into product design and manufacturing, we adopt the definition for cloud-based design and manufacturing as proposed by Wu and co-authors [7]: “Cloud-Based Design and Manufacturing refers to a product realization model that enables collective open innovation and rapid product development with minimum costs through a social networking and negotiation platform between service providers and consumers. It is a type of parallel and distributed system consisting of a collection of inter-connected physical and virtualized service pools of design and manufacturing resources (e.g.: parts, assemblies, CAD/CAM tools) as well as intelligent search capabilities for design and manufacturing solutions.”

In this paper, we present PDSIDES Version 2.0, namely, Cloud-Based PDSIDES (CB-PDSIDES). CB-PDSIDES is currently developed to run in a virtual machine on Google Cloud Computing Engine. The platform PDSIDES integrated with the cloud has in its core a decision support gene. We address two types of decision support genes via CB-PDSIDES: compromise and selection decision support problems (known as cDSP and sDSP, respectively) [19, 20]. The execution of decision support genes on a computer is foundational to the platform to realize complex engineered systems. The developed ontologies for selection decision, compromise decision, and hierarchical coupled decisions are implemented in CB-PDSIDES for representing decision workflows. In Fig. 1, we illustrate the concept of CB-PDSIDES with the cDSP construct at its core as the fundamental decision support construct. The problem-specific information (declarative knowledge) is captured using the cDSP via the keywords Given, Find, Satisfy, and Minimize. Procedural knowledge is associated with how the information transformation is carried out and details how the transformation is executed via a decision workflow or decision network. This is captured via decision-based design templates like process template, sDSP template, cDSP template, surrogate modeling template, design space exploration template, and robust design template [21]. The analysis codes and simulations associated with the problem framed are also communicated to the decision support construct, see Fig. 1. Surrogate modeling techniques and tools are available in the platform via the surrogate modeling template to support the designer in managing the complexity and coming up with reduced order models (see [21] for details). Post-processing tools like ternary plots and contour plots that are automated with rules are available via the design space exploration template to help the designer easily explore the solution space and identify robust solution regions of interest by managing uncertainty (see [21] for details). Frameworks and design methods anchored in the decision-based design paradigm that incorporates the decision support genes (cDSP or sDSP) are incorporated in the cloud-based platform to support the designer to formulate and execute design problems systematically. These include Robust Concept Exploration Method (RCEM) [22], Product Platform Concept Exploration Method (PPCEM) [23], Inductive Design Exploration Method (IDEM) [24], Concept Exploration Framework (CEF) [25], Goal-oriented Inverse Design (GoID) Method [26], and GoID Method with robustness [9, 27]. Integrated knowledge, i.e., integration of declarative and procedural knowledge, is captured via the design methods and frameworks available in the platform. The computational solvers associated with the execution of problem formulated like DSIDES (Decision Support in the Design of Engineering Systems) [28] for cDSP construct are available to be accessed via the cloud. Collaborating designers can access the cloud-based platform from different parts of the world to share information and formulate design problems worthy of further exploration. The knowledge associated is captured and stored in CB-PDSIDES and can be retrieved and shared instantly via cloud with the collaborators depending on the design requirements. The issues of collaboration and information sharing are also addressed as collaboration and communication is key in the Cloud-Based PDSIDES. This is addressed via paradigms like crowd-sourcing, mass collaboration, and social product development [3].

In Fig. 2, we show the architecture of CB-PDSIDES. The computing architecture for CB-PDSIDES follows the architecture of cloud-based design and manufacturing systems proposed by Wu and co-authors [29]. The architecture of CB-PDSIDES includes five layers: (i) user layer, (ii) web portal layer, (iii) logic layer, (iv) virtual layer, and (v) physical layer.

Since CB-PDSIDES is deployed in the cloud, users can gain access to CB-PDSIDES via PCs and smart phones over the internet. The web portal layer of CB-PDSIDES includes the user interaction GUI for accessing the design templates available. The user interaction GUI includes the following: template searching and browsing GUI which are designed for locating the required DSP templates and presenting them; template creating and editing GUI which are designed based on the DSP template structures for the purpose of instantiation and modification of the DSP templates; and the template execution and analysis GUI which are designed for executing DSP templates and performing post-solution analysis. The GUI is allowed to communicate with the logic layer of CB-PDSIDES by a request-response mode using the Hyper Text Transfer Protocol (HTTP). The logic layer of CB-PDSIDES includes five main parts, namely, Response Server, Knowledge Base, JESS Reasoner, DSIDES, and MATLAB. The Response Server is the central “brain” that integrates the other four parts for responding to requests. The Response Sever itself has five components including a search engine, an instance interpreter, a consistency checker, and a problem solver. The instance interpreter is for interpreting the data collected from the template creators (or editors) and formatting it into DSP template instances according to the DSP ontologies. The generated template instances and module instances are stored in the Knowledge Base. The search engine is connected to the Knowledge Base to provide ontological semantic-based knowledge retrieval. Consistency checking is facilitated through a consistency checker together with the JESS Reasoner—the Rule Engine for the JavaTM Platform, which provides rule-based intelligence inference. The problem solver is connected to DSIDES for solving the DSPs. DSIDES (Decision Support in the Design of Engineering Systems) is a tailored computational environment that supports the execution of the decision support problem templates. DSIDES is invoked when a template executer executes a DSP template. The result analyzer is included to help users especially template implementers analyze the results produced by the problem solver, DSIDES. MATLAB and its features support data visualization tools such as ternary plots and scatter plots for visualizing the DSP results and further carry out solution space exploration. Therefore, MATLAB and its features for visualization and solution space exploration are integrated to CB-PDSIDES to support post-processing. The virtual layer and physical layer in CB-PDSIDES support data storage and retrieval for multiple users who are connected via cloud thereby establishing a networked collaborative environment. As noted by Wu and co-authors [29], through virtualization, the computing resources of CB-PDSIDES like the DSP templates, solvers, post-processing resources, and data storage are separated from physical computing hardware and reallocated dynamically to the different applications based on the needs of users. Through this unified computing architecture of CB-PDSIDES, we are able to support multiple tenants through a single instance of the platform PDSIDES, defined as multi-tenancy [29]. These features of the computing architecture of Cloud-Based PDSIDES differentiates it from the other web-based services. In the next section, we present some of the core functionalities offered by CB-PDSIDES in ICME context for integrated materials and products realization.

Core functionality 1 (CF1) is the capability of CB-PDSIDES to support a designer in designing new design workflows by reusing past knowledge from similar design problems. Specifically,

CF1 is possible using CB-PDSIDES by means of an ontology that provides a common vocabulary for representing domain-specific knowledge. The ontologies for representing the knowledge captured via the decision support genes, cDSP and sDSP, have been developed and form the foundations for CB-PDSIDES (see [] for details). A PEI-X (Phase-Event-Information-X) ontology for meta-design is included in CB-PDSIDES for designing decision workflows. Details regarding the ontologies developed are available in [14, 18, 21].

The focus from the ICME context is on concurrently exploring the design space of both the products and the materials and narrow the set of possible options in the shortest possible time with minimum expense. Hence, instead of exploring the complete design space from first principles using detailed models, the focus is on exploring simplified models that are good enough to compare different design alternatives. Additionally, the notion of designing the workflows by reusing past knowledge from similar design problems is important because of the following:

Moreover, the needs for accurate information depend on whether the goal is to narrow down the alternatives to a specific class of materials (i.e., during early design phase) and products or to design the composition and structure of a specific material system (i.e., during the later stages of design). To support the need to generate information at variable fidelities during the design process, the following features are offered via CF1 by CB-PDSIDES:

CF2 is the capability of CB-PDSIDES to support designers in design collaborations. CF2 is possible using CB-PDSIDES through a Secure Co-Design (SCD) Framework that uses information theory [30, 31] and game theory [32, 33] protocols to identify co-design solutions while preserving confidentiality of the information shared between the participating design collaborators.

From the ICME context, this could be a scenario where components are designed by one organization and materials are designed by another organization where information and knowledge sharing needs to be managed. An example of this could be a product-level cDSP formulated by product designers and a material-level cDSP formulated by materials scientists. Let us assume that the material scientists have proprietary models that they do not wish to share (explicitly or implicitly) with the product designers. In a similar sense, the product designers do not wish to share all the information in the product-level cDSP with material scientists. However, both parties would like to jointly design the product and the material and are connected to each other via cloud. The collaborative nature of the design process induces additional design issues from ICME context to be addressed for the management of design workflows, such as the following:

There are several methods for simulating various aspects of materials’ manufacture and product design. However, it can be very costly and time-consuming to compute and re-compute these simulations in the process of design, especially in the early stages of design where it is desirable to explore a wide range of options rather than developing detailed designs. For situations like this, there are several ways of developing surrogate models (metamodels) which rapidly provide design information; each of these will give metamodels of different degrees of accuracy at different costs which may be used at different stages of design (see [34, 35] for details). There is thus a need to assess the benefits of using different metamodels in different stages of design and compare these with the costs of developing these metamodels. Using metamodels of increasing fidelity in a design process is one way of exploring the design space; an alternative way is by using robust design with decreasing bands of robustness. The advantages and limitations of each of these approaches has to be considered. To support the need to reduce computational cost and manage complexity, CB-PDSIDES offers the following capabilities:

This functionality is addressed by two modules in design space exploration (DSE) process template ontology in CB-PDSIDES (see [21] for details). The DSE process template has three sub-templates: problem model (PM), compromise decision support problem (cDSP), and post solution analysis (PSA). The problem model (PM) sub-template in DSE process template has two modules: theoretical and empirical model and surrogate model. The theoretical and empirical model module is used to capture the information and knowledge associated with already existing and available material and product models. An example for this is capturing the information and knowledge associated with an existing constitutive material model in literature that establishes stress as a function of strain, strain rate, temperature, and other material internal variables. The surrogate model module is used when there is a need to develop reduced order models or meta-models that captures the relationships between responses and the corresponding factors. Template instances of the surrogate models developed are stored as new knowledge in CB-PDSIDES to facilitate reuse. Collaborating designers are able to use the cloud to develop, share, and use models to formulate design problems using CB-PDSIDES.

CF4 is the capability of a designer to use CB-PDSIDES in exploring the effects of changing the ways in which templates are integrated on the outcomes of the design workflows. This is achieved in a modular fashion by means of a 3-P information model proposed by Panchal and co-authors [30]. The three P’s refer to the key design information elements—product, problem, and process. Each of these is defined in an independent modular fashion. Designing a product is possible by using different types of un-instantiated problem and process templates. Examples of different possibilities in problem template includes Archimedean cDSP formulation [36], preemptive cDSP formulation [19], Robust Design Type I formulation [37], Robust Design Type II formulation [37], Type I, II, and III Robust Design formulation [38], Robust Design Type IV formulation [24], and traditional optimization. Examples of different possibilities in process template includes point-based iterative search, sequential design, set-based design, RCEM or CEF using response surface models [22], and goal-oriented inverse design of process chains [26]. Examples of the different product templates include design of pressure vessel [15], design of hot rod after rolling [39], design of slab after casting [40], and design of multiscale materials and products [41]. The benefit associated with this functionality from ICME context is the capability to realize configurations of different problems with different processes and applying these for a variety of product design scenarios. This leads faster product realization schemes in a cost-efficient manner as needed for ICME. Using the cloud, participating material scientists and designers are able to collaborate in the integration of templates for product development depending on their expertise—in problem, product, and/or process information elements.

Design and solution space exploration by considering system uncertainty is essential for the model-based realization of complex material/product systems via ICME. As discussed in “Frame of Reference,” the models that are available are typically incomplete, inaccurate, and not of equal fidelity. Hence, seeking single point optimum solutions is typically not valid as these optimum solutions no more hold if any variations occurs—which is bound to happen with a complex material/product system. This necessitates the need for systematic design and solution space exploration to identify satisficing solutions that perform well and are relatively insensitive to the uncertainty present in the system—defined as robust solutions. Using the platform CB-PDSIDES, designers can collaborate from different parts of the world in formulating robust design problems and exploring the design and solution space using rules defined in the platform to identify solutions that are relatively insensitive to these variations. The exploration of the solution space provides designers with knowledge to refine or improve the model especially at early stages of design. This functionality is supported by the post solution analysis (PSA) sub-template within the design space exploration (DSE) template of CB-PDSIDES (see [21] for details). The three modules Weight Sensitivity Analysis, Constraint Sensitivity Analysis, and Additional Requirements Analysis that form part of PSA sub-template of CB-PDSIDES are used depending on the requirements of the problem to explore and visualize the solution space and identify solutions that are relatively insensitive to uncertainty. The designers are able to formulate a robust design problem by instantiating a Type I, II, III, or IV Robust Design Problem template in CB-PDSIDES.

A key goal in design is the proper communication of design process. Wu and co-authors [7] identify the key issue of fully understanding a complex design process in order to improve design communication. The key issue includes design tasks that need to be completed, the source for specific information that is needed for design, the individual to be contacted for the right information, the extent of distortion in the information available, and the extent to which the distorted information affects design. As discussed in “Frame of Reference,” traditional product design paradigms are limited as design communication is a one-way mapping in a linear sequence across design phases/domains. PDSIDES in the cloud settings will improve design communication through multiple information channels facilitated by cloud. This will allow for information flow in multiple directions facilitating dynamic changes during product development and instant communication between the different design domains. Assume that each stage of product realization starting from function to final manufacturable product descriptions involves distributed designers. PDSIDES facilitates a decision network where information is shared in a one-way fashion. CB-PDSIDES however can facilitate collaboration and can result in a two-way and multi-way network for product realization where the distributed designers are connected through the cloud. In the next section, we demonstrate the core functionalities discussed via an industry-inspired problem.

From the ICME context, we are interested in designing the next generation of cyber-physical-product/material systems. Design of products, sub-components, and materials with improved system-level performance goals is the major focus here. One application area of focus is the design of American football helmets. The goal-oriented inverse design method and Concept Exploration Framework (CEF) incorporated into the CB-PDSIDES platform are used to design two components (composite shell and foam liner) of an American football helmet by Fonville and co-authors [42]. The performance goals defined for the helmet in this problem include dissipation of impact energy, mitigation of stress waves, and minimization of helmet weight (see [42]). The larger vision here is to design products, sub-components, and materials using multi-scale modeling efforts and system-based robust design techniques so as to achieve the integrated design exploration of multiscale, multifunctional materials and products.

One other application of focus for CB-PDSIDES from the ICME domain is in cyber-physical design. An application example is the design of steel manufacturing process chain discussed in this paper. PREMΛP—Platform for Realization of Engineered Materials and Products—is developed by TCS Research, Pune, as a comprehensive IT platform that facilitates the integration of models, knowledge, and data for designing both the material and the product [43]. The platform PREMΛP is developed for different types of users like expert users, non-expert end-user, and for researchers. Gautham and co-authors [43] define the domain-independent and domain-dependent components of PREMΛP. The CB-PDSIDES presented in this paper has the potential to support PREMΛP with several of its components. Based on the key functionalities proposed for CB-PDSIDES, it is envisioned that CB-PDSIDES can support PREMΛP in robust design and MDO, decision support, knowledge engineering, guided experimentation, product design, and product performance.

Application of CB-PDSIDES to cyber-physical-manufacturing systems is explored from the context of Industry 4.0. Smart manufacturing systems design in the era of Industry 4.0 demand a need to address the distributed and networked nature of manufacturing processes and their associated products. There is a need to design manufacturing systems that facilitate seamless data, information, and knowledge sharing between the different physical, cyber components of the system and the stakeholders (customers, suppliers, manufacturers, etc.) involved. Milisavljevic-Syed and co-authors [44] present a computational framework for design of dynamic management of such networked manufacturing systems from the context of Industry 4.0. It is envisioned to incorporate the framework and associated constructs to CB-PDSIDES with possible applications of design for dynamic management in steel manufacturing, additive manufacturing, health care systems, etc.

CB-PDSIDES for automated micro-enterprise design and analysis for sustainable rural development of villages in India is also explored as an application to cyber-physical-social systems. Three constructs that support this application are (i) village-level baseline sustainability index, (ii) dilemma triangle, and (iii) village-level system dynamics, proposed by Yadav and co-authors [45]. Yadav and co-authors [45] present a computational framework incorporating these constructs to facilitate dialog between the stakeholders involved: corporate social responsibility (CSR) investors and social entrepreneurs. These stakeholders are geographically dispersed, and an increase in the number of stakeholders demands a need for them to effectively communicate and collaborate in order to come up with sustainable solutions (micro-enterprises) worthy of further investment. We envision incorporating the framework and constructs developed by Yadav and co-authors in CB-PDSIDES to facilitate collaboration and open innovation between these stakeholders. Using CB-PDSIDES, the stakeholders will be able to direct attention to issues and challenges that are typically ignored or missed while solving social wicked problems.

All these applications discussed are work in progress in the International Systems Realization Partnership between the Institute for Industrial Engineering at The Beijing Institute of Technology, The Systems Realization Laboratory at The University of Oklahoma, and the Design Engineering Laboratory at Purdue. Other collaborating institutions include Center for Advanced Vehicular Systems at Mississippi State University, Liberty University, SunMoksha, Tata Consultancy Research Pune, and the University of Liverpool.