3.1. Gear 1: Strategy

Manufacturing organizations have started to appreciate and embrace Industry 4.0 realizing the potential it can offer in terms of improving efficiencies, cost savings and responsiveness to the market demands ^[17]. However, it can be confusing on how it can be achieved in practice especially with limited case studies available and the journey could vary from one organization to another ^[6][21][28]. To achieve the goal towards smart manufacturing, the proposed models suggest building a strategy to assist manufacturing SMEs throughout their journey. Figure 4shows four main areas that should be considered in this stage.

Figure 4. Gear 1—strategy stage.

Firstly, a business case should be in place with clear vision, goals and objectives of what is to be achieve that best suits the organization’s business model. This can be done by prioritizing the areas in which it will offer the most benefit and long-term Return-On Investment (ROI) ^[9]. Recent surveys from the aerospace industry suggest that the majority expect Industry 4.0 investments to payback within two years ^[25][38]. Early and fast adoption via launch and lean approach could also be significant for competitive advantage and cultural change over late adopters ^[25]. Manufacturing organizations within the same industry might share similar drives and goals, for example, the aerospace industry common drives are to increase productivity as demand grow, reduce the cost of manufacturing for competitiveness and to add value to the products and services they offer ^[8][38]. Every manufacturing company can have a different starting point, and a bespoke digital transformation journey. It’s important to identify a specific starting point, understand where the value to be created and begin the journey ^[39].

Secondly, a key to a successful journey is the role of leadership. The challenge in the manufacturing industry is that leadership is not at the required level to deliver an optimum strategy for faster technology adoption ^[28] because of lack of available examples or case studies from which leaders can inspired and take lead from. To overcome the leadership challenges, digital transition should be a strategic investment—Possibly written in company’s annual strategic document-to ensure an informed investment to get the confidence on the ROI. Leaders should also consider the wider technical infrastructure and skills needed to manage and accommodate the technologies in-house and what is needed to source them. Managing cultural change and getting the buy-in should be driven from the top with clear communication strategy on where the organization is heading and why they are doing it in order to bring the whole organization onboard throughout the journey ^[8][28][29].

Thirdly, is to appoint a champion. Due to the nature of an Industry 4.0 project, i.e., being complex, requires multi-disciplinary skills and cross communication between different departments that often operate independently (e.g., IT and operations), a champion with the right leadership and technical skill set is needed that can act as a digital officer and a project manager whose responsibilities include: (i) overseeing the strategy; (ii) facilitating activities between various departments; (iii) managing the project implementation and technology adoption; (iv) constantly updating and communicating the ROI as well as success metrices to leadership; and (v) establish vison of future potential projects and applications to scale along the journey ^[40].

Finally, foster collaboration and plan them while developing the strategy. Industry 4.0 projects are complex as they touch on many inter-disciplinary areas (e.g., integration of sensors, IT, analytics, etc.). The level of technical readiness, maturity and skills of the company workforce should be assessed to identify the relevant support needed. Collaboration and partnership with academia, technology hubs and technology providers at different stages of the roadmap should be considered. Specialist skills and development support enable quick wins and faster adoption, de-risking failures and investments ^[28][40][41].

3.2. Gear 2: Connectivity

The connectivity stage of the roadmap is the most essential stage because it serves as the foundation for other stages to build upon. Figure 5 shows the three sub-categories that should be considered and the three key enabling technologies associated in this stage.

3.3. Gear 3: Integration

This stage is associated with the integration of IT and OT systems in manufacturing that often operate independently, the former manages business applications from the front office and the latter keeps the plant operations run smoothly. Integration of IT and OT are the backbone of the smart manufacturing architecture which have been addressed by various Industry 4.0 smart manufacturing frameworks, standards and reports ^{[1][5][6][8][9][10][15][33][62]}.

Figure 7 highlights vertical and horizontal integration requirements from an operational perspective that should be considered in this stage. Vertical integration aims to tight together all the operation layers within the manufacturing organization keeping as much of the value chain in-house as possible. The horizontal integration on the other hand aims to connect networks of cyber-physical and enterprise systems presents within the production facilities and the entire supply chain to establish partnerships building an end-to-end value chain ^[63]. In this stage, vertical integration is discussed in detail as its considered essential for systems and data integration between IT and OT, in order to provide agility in finding patterns for finer control of the manufacturing process ^[4]. Horizontal integration is discussed briefly in stage 6 when scalability adds additional value to the connectivity and integration architecture of the smart factory.

Figure 8 presents an example of the international standard ANSI/ISA-95 traditional manufacturing operation layout ^[64]. It encompasses five traditional operational levels (0–4) where manufacturing information exists side-by-side together with some examples of manufacturing software applications used in each level. It is widely developed and applied in the manufacturing industry to provide the terminology for clarifying application functionality and the way information is used.

IT systems in the manufacturing industry have been traditionally developed to collect data and manage work processes within manufacturing, to meet specific functional areas within each operational level ^[65]. These systems have been purposely built to resolve specific set of issues and often operate independently. They have difficulty obtaining data across operational levels and they can’t scale or sustain with the increasing complexity of manufacturers needs ^[65]. For example, enterprise resource planning (ERP) systems used in the business operation level play a major role in connecting day-to-day activities across the manufacturing facility, providing essential insights into operations and the manufacturing process integrated into one database ^[66]. In the operational management level, manufacturing operation management (MOM) systems and manufacturing execution systems (MES) operates to bridge the gap between the enterprise business level and the automation levels but there are limitations and complexities associated with connectivity and vertical integration of real-time production data ^[33][67]. They are still heavily dependent operators’ inputs and lack data processing analytics capabilities.

Connectivity, vertical integration and synchronization between the manufacturing operations levels are essential for greater insight into the production process for rapid decisions making, corrective actions to be taken and to enable optimization. This can have a significant improvement in business and financial performance making it one of the challenging applications and cornerstones of Industry 4.0 vision for smart manufacturing ^[68][69]. Recent reports have shown that only few digitally capable advanced manufacturing companies have fully integrated and correlated plant data with enterprise applications since a more advanced connectivity and integration tools are needed ^[68][67].

There are number of key challenges associated with the integration stage. They include: (i) interoperability (breaking down the silos in the factory floor machines and equipment’s that speak different languages depending on the communication protocols set by the vendors); (ii) data security (privacy issues with external sources and third parties in the horizontal integration as it requires data to be kept secured and accessed on a need to know basis); (iii) scalability (easily scaling the IT infrastructure by shifting IT systems operation to the cloud when data volume and velocity significantly increase with time); and (iv) adopting cost-effective IT applications (with strong orchestration capability to handle the integration complexity as well as providing end-to-end visibility with robust tools for data analytics to be performed) ^[70][63][71].

The IT technology market is already providing solutions to overcome the complexities of IT and OT integration to solve the challenges of interoperability (using open and multiple standards and protocols for communication), data security (using best practice CS standards and protocols) and scalability (via cloud ready solution) ^{[44][52][67][69][71]}. Table A2 in the Appendix A shows a list of leading Industry 4.0 ready IT platforms used in the manufacturing industry that use best practices connectivity and integration architecture that combine elements of IIoT and MOM/MES. These platforms do not completely replace or rebuild long established systems but acts as a new layer to the infrastructure orchestrating the new way of Industry 4.0 in smart manufacturing. This involves sourcing data from anywhere with minimal connectivity and integration costs together with having analytical capabilities to deliver insights via digital dashboards that can be accessed from anywhere including mobile devices, augmented and virtual reality displays.

Manufacturing companies should carefully consider when implementing an IIoT system whether to buy off-the-shelf solutions or build an in-house solution taking into account the skills needed to build the technology stack, the scalable capability of solution, the availability of resources, technical skills and time ^[4][72]. Industrial research reports find that most successful IIoT projects use off-the-shelf solution as the basis of their operation can get a quick ROI on the technology implementation ^[4][67][72].

A strategy should be in place for choosing an IIoT solution to suite the company’s appetite. Some of the considerations include the additional capability of the platform, the additional IT infrastructure and hardware needed, the cost of using the platform, the customization flexibility, the technical skills required to use and manage the platform and the ability to scale in the future.

IIoT platforms with built-in MOM/MES applications are a viable option. They are capable to seamlessly integrate with enterprise and industrial automation applications ^[66] and are accompanied with the elements of the IIoT stack (e.g., analytics tools, IIoT gateways connectivity, cloud scalability) ^[8]. Case studies in ^[73][74] shows examples of manufacturers implementation of such option. Other option combines industrial code-less IT systems with built-in software bridges, modules and communication protocols that are capable of connecting and integrating data across the manufacturing operation layers ^[52]. Examples of such IT systems are OPC Router ^[75] and Kepserver ^[76]. Another option is to adopt a full-fledged IIoT platform to seamlessly connect and integrate all layers of the manufacturing operation, enhanced with additional manufacturing operation apps ^[59]. Case studies in ^[77][78] shows manufacturers implementation of such option. Embracing an ecosystem of solutions that work in harmony enable a successful system level integration ^[67]. All three strategies have the potential to combine data from different manufacturing operation levels enabling capabilities to move to the next stages of the roadmap.

Figure 9 shows an illustration of a smart manufacturing integration architecture that combines elements of the manufacturing operational levels applications with elements of the communication layer of the IIoT stack bridging the gap between IT and OT resulting in a connected and integrated smart factory.

Figure 9. Manufacturing operation levels and IIoT integration architecture. (Author’s creation).

3.4. Gear 4: Analytics

This stage is associated with analyzing data collected from different sources of the manufacturing operation layers. Big Data refers to the vast amount of data generated due to the volume, variety, velocity, variability and value ^[79]. Big Data is categorized as smart data when collected in a structured manner, condensed, processed and particularly analyzed in a way that turns it into an actionable information. Data analytics create value in manufacturing by providing a diagnostic approach to enable root-cause problem solving and guidance to reduce deviations that occur during manufacturing. Big Data is categorized in three ways: (i) structured data (data that comes from traditional database and equipment); (ii) time series data (continuously changing and event-based data); and (iii) unstructured data (other kind of data and information not typically used in manufacturing, e.g., weather, videos, etc.) ^[27][80].

The challenges of dealing with the vast amount of data include data handling complexity, data volume, speed of the network and bandwidth to process such large amount of data. To reduce the handling complexity and network bandwidth issues/limitations, data should be initially extracted, handled, cleansed, time stamped and processed at the edge (e.g., in the IIoT gateway/machine control level). This is to produce relevant information before transferring it to the fog (IIoT platforms/operation management level applications) for further processing and detailed analysis ^[81].

The traditional view of the analytics framework, that drives improvements of industrial operations, is divided into four types: descriptive, diagnostics, predictive and perspective ^[9][80]. Descriptive analytics are a set of metrics that describe the event (what happened). They are referred to as key performance indicators (KPIs). These KPIs are represented by mathematical calculations to provide insights into the performance of manufacturing operations ^[82]. Diagnostics, predictive and perspective are analytics that uncover questions such as, why things happened, what would happen as a result and what action to take in such instances. Descriptive, diagnostic and predictive analytics are widely used as analytical tools in the manufacturing industry to drive operational improvements ^[80]. Manufacturers, largely SMEs, still rely on common KPIs to gauge the health of their business ^[27].

Figure 10 shows common descriptive and diagnostics analytical applications (e.g., Productivity KPI metrices, condition monitoring, performance analysis) for this stage along with the technology jigsaw map showing Big Data and analytics tools to be the fundamental technologies in this stage. Predictive and perspective analytics and their applications are described in the next stage.

Figure 10. Gear 4—analytics stage and Industry 4.0 technology jigsaw (Primary technologies represented in color, secondary technologies represented in grey).

Amongst the most common and popular productivity KPIs that the manufacturing organizations rely on (for managing operations) are mainly focused on financial business performance (e.g., manufacturing cost per unit), efficiency focused metrices (e.g., factory efficiency, overall equipment effectiveness (OEE)) and quality related metrics (e.g., scrap rate, yield) ^[83]. However, more from the acquired data can be achieved by implementing a range of KPIs that allow for the assessment and tracking of the manufacturing process as well as success evaluation in relation to the goals and objectives. Reporting of KPIs can also be improved using real-time digital visualization dashboards built within MOM applications, IIoT platforms or business intelligence (BI) tools.

Asset condition monitoring is a snapshot of the health conditions of the assets in the factory (e.g., manufacturing machine). Tracking changes of the status, performance, utilization, etc. of manufacturing machines enable significant change to be easily spotted once occurred. Historical records of the snapshots collected over time from various machines in the production line enable comparison and performance analysis to be performed ^[7]. For example, machines utilization history, maintenance history, failures modes etc. can be captured and compared with its peers to identify trends and behaviours related to the manufacturing process. This is valuable for diagnostic analytics to enable manufactures understand why things happened.

Industrial research reports in ^[27][80] recommend manufacturing SMEs to consider reviewing KPIs already being used with industry standards to ensure most important metrics are captured that drive operational performance. They also recommend KPIs tracking, their transformation into a continuous development program for operational excellence and preparation for the move into predictive and perspective analytics.

3.5. Gear 5: AI (Artificial Intelligence)

This stage is associated with the advanced data analytics and tools used to provide new answers to questions raised from the data analysis stage to add real and unexpected value to the manufacturing industry. It is based on using Big Data, predictive and perspective-based analytics and artificial intelligence-based analytics.

AI is the ability of the computer to think and learn itself. Big data, intelligent algorithm and computer systems drive the AI process to enhance machines and people through digital capabilities such as perception, reasoning, learning and autonomous decision-making ^[84]. In the manufacturing industry, AI is seen as the brain of the Industry 4.0 transformation while other digital technologies provide the muscles to drive the transition from automation to autonomy.

Key driving factors to implement AI in manufacturing includes: (i) to improve cost savings; (ii) to transform operations; (iii) to provide better customer service; and (iv) to create better workplace environment ^[85][86][87]. AI also has a potential impact on the economy leading to a stronger and improved economical prospect as a result of a more efficient, sustainable and competitive manufacturing ^[3].

Applications of AI and the use cases vary in manufacturing. Predictive maintenance of machines and tools is one of the key values created, which has the potential to reduce repairing costs by 12% ^[87]. AI-based learning algorithms spot trends in the data for early warnings and indications of possible failures and breakdown. This allows maintenance to be scheduled and intervened rather than being dependent on periodic checks, enabling a proactive behaviour instead of a reactive behaviour, which leads to a more reliable and sustainable production line.

An example of an AI predictive maintenance application in the aerospace industry is used by the British engine manufacturers Rolls Royce. The IoT capabilities and advanced analytical tools of Rolls Royce engines in the aircrafts are capable of predicting wear and tear in the engine so that an intervened repair and maintenance awaits as the airplane lands in order to maximize the aircraft availability ^[88]. Another application of AI in the aircraft industry is the Airbus Skywise platform ^[89] empowered by machine learning algorithms to enable insights into the data, collected from various sources of the aircraft for maintenance and repair operations. They also provide insights into operational efficiencies optimize the parts usage, reduce aircraft downtime and cut down service costs.

A manufacturing application of predictive maintenance is presented in the case study by Lee et al. in ^[90] where an intelligent spindle monitoring system was developed to monitor and predict the performance of the tool spindle of a CNC machine minimizing maintenance costs and optimizing product quality. AI application in manufacturing also plays crucial role in empowering workforce with a baseline knowledge to improve process efficiency and productivity. AI based algorithms can provide workforce with real-time recommendation(s) to act upon especially for young inexperienced workforce. They can also autonomously act to address raised issues making machines to adjust themselves in order to optimize quality or energy efficiency during production operations ^[91]. Other use cases of AI in manufacturing include but are not limited to: (i) improvements in quality inspection using images processing and recognition techniques for process optimization and scrap reduction ^[85][87]; (ii) in safety monitoring and control using AI-based self-learning models embedded within manufacturing automation systems (e.g., robots). These systems learn from prior experience and human interventions and react to unforeseen situations, resulting in taking pressure off people, reducing human error and improving workplace health and safety ^[85]; (iii) production scheduling using AI, combined with mathematical optimization tools to plan, schedule and optimize capacity planning ^[92].

Although, many manufacturing companies are making significant steps in AI adoption, there are still challenges that form barriers to AI uptake. A recent global survey reported by plutoshift ^[93] indicates that only 17% of manufacturing organizations are in full implementation stages of AI. These challenges are normally associated with the lack of strategy. The nature of an AI-based system is that it needs to be continuously trained, monitored and evaluated while guarding against bias, privacy violations and safety concerns; therefore, an AI strategy should be in place. The Microsoft AI Maturity Model ^[85] is a good example that highlight AI development. It consists of four stages namely Foundation, Approach, Aspiration, Maturity and is based upon organization’s assessment and maturity level to guide adoption of the right kind of AI tool(s) at the right place and time.

Figure 11 shows a simple process for manufacturing companies to use as a guide to deploy an AI strategy. The aim is to enable predictive and perspective analytics that can be practically achieved. It consists of five steps and utilizes six Industry 4.0 technologies. The first step is to identify a business case that is aligned with the business strategy (e.g., to maintain a reliable and sustainable production line in order to improve customer service). The second step is to choose a suitable analytical application to meet the business case objective (e.g., predictive maintenance of machines in the production line). The third step is to consider how to best build and implement the application. Examples include (i) the type of analytical tools needed to build the application (e.g., Software service/IIoT platform) (ii) the type of Big Data needed (e.g., structured, unstructured, time series data) (iii) the location of storage and the computing power needed (e.g., local, cloud), (iv) the analysis methods to use for the identified application (e.g., machine learning, deep learning techniques) (v) the skills needed to build the application and (vi) the cost of implementation (In-house development or outsourcing). All these should be considered while taking into account the culture and capabilities that match the maturity of the company. The fourth step is to have an adequate alerting, visualization and reporting mechanism in place to enable people to make decisions and take actions based upon the information produced. Machines and robots can also be further enabled to execute actions themselves. The process can allow them to self-adapt to the control commands based on the information provided by the AI application. This subsequently enables autonomy and the application of CPS in manufacturing. The fifth step is to have a continuous feedback and configuration control to optimize the AI application, the preventative actions and the decisions being taken.

Figure 11. Gear 5—AI stage and Industry 4.0 technology jigsaw (primary technologies represented in color, secondary technologies represented in grey).

Main challenge in this stage is associated with data. Data availability, data quality, adequate connectivity and data infrastructure are needed to embrace analytics and fully utilize AI applications ^[93]. These challenges can be addressed by using robust IIoT platforms/analytics applications that are capable of data mining from anywhere, perform necessary computing and providing adequate visualization of information that translate data into actionable information ^[85][68].

Other challenges of adopting AI are associated with cultural change and new technology adoption. These include perceived fears of workforce around technology being too difficult to use, lack of trust, or often the fear that it will take jobs away. However, according to HPE survey in ^[86], it is predicted that AI-created jobs will balance or outweigh those made redundant. Therefore, leadership from the top should have continuous communication strategy in place to overcome such issues. There is also a lack of data science knowledge and AI skills in the manufacturing industry ^[85][86][87], therefore, professional development of the internal staff, hiring of technically skilled people or collaboration with external expertise in the field could possibly overcome these issues.

3.6. Gear 6: Scale

The proposed “Six-Gears Roadmap” is anticipated and envisioned to be dynamic and continuous. The sixth stage in the process is thus related to the scalability stage. It is designed to increase the scale and scope of the application within each stage in the roadmap by capitalizing upon the capacity of the Industry 4.0 digital technologies.

Figure 12 shows an overview of different options available to scale the stages of the roadmap together with showcasing key building blocks of the technologies associated with each stage. The roadmap can be scaled horizontally and/or vertically to show the different opportunities available for manufacturing SMEs to capitalize upon. The horizontal path represents the options to scale any stage of the roadmap together with an additional option that represents other key Industry 4.0 technologies that have potential in smart manufacturing. The vertical path on the other hand represents the option with further deep dive into each stage of the roadmap.