Foundational Architecture for Industry 4.0
Many industrial and manufacturing enterprises have started Industry 4.0 (I4.0) initiatives to increase agility, improve operational performance and increase revenue from the raft of new digital products and services now available. These initiatives have brought some success but the transition to the Industry 4.0 vision of smart, digitally connected and autonomous production lines, factories and supply-chains is still at the initial stages.
It is very important for these enterprises to keep progressing on their journey as it will be difficult to catch up with the entrenched digital manufacturing leaders at later stages. To enable progression towards I4.0 they must build foundational enterprise application, data and computing architectures which can evolve with the new vision of the digital enterprise. Some of the key elements/initiatives for establishing foundational architecture are described here.
- Manufacturing Execution/ Operations system - There is a view that the MES systems need time-consuming implementations, benefits are hard to quantify, are inflexible and not needed in the IIoT world. The reality is that MES systems or some flavor (in form of IIoT apps for specific functionality) are needed as a foundation for high operational performance, product traceability, real-time visibility into shop-floor operations and the contextualization of shop-floor data being collected for analysis for further operational improvements. The enterprises should select a MES system which is suited to their industry and manufacturing process type, is modular, delivers substantial out of the box functionality with basic configuration, is extendable, can scale across the enterprise beyond individual factories and has a product roadmap to continuously benefit from IIoT developments. The MES should be interfaced with ERP systems based on the ISA-95 standards and with PLC/DCS on shop-floor using OPC-UA.
- Product Engineering and Innovation - Customers today expect smart connected products, customized to their current needs at acceptable costs, which are also capable of improving with their evolving requirements. The traditional PLM systems used by companies now need to mature as platforms for product innovation. They need to include additional capabilities such as model-based systems engineering and integration of model-based simulations with the actual IIoT data from deployed products/ plant assets. This is required for building digital twin capability and better integration with shop-floor and supply chains for capturing product information across the digital thread. The manufacturing enterprises need to select/update their PLM /PDM systems in line with these trends.
- Data and Application integration – A coarse-grained business integration layer can be implemented which can serve the needs of any apps or API/services provided by the enterprise. The applications should be enabled to support REST/SOA services based interfaces. Additionally, event driven messaging architecture using publish/subscribe mechanisms should be implemented for business process workflows. A canonical data model for common business objects (could be based on industry specific reference model) should be used within the enterprise and partners/suppliers so that data translation is minimized.
- Distributed Analytics and Decision-making infrastructure- Most of the manufacturing enterprises will have historical databases and data warehouses for analysis of shop-floor and Enterprise IT data. This established analytics infrastructure should then be combined with new data sources (e.g. - IIoT, social media, weather) using data lakes on Cloud infrastructure for big data analytics and machine learning, to achieve enterprise wide process improvements. This will be in addition to the real-time monitoring and control needed at plant devices/ equipment (Edge) and at PLC/SCADA systems which can also serve as IoT Gateways for sending shop-floor data to Cloud based big-data analytics infrastructures. The insights from analytics infrastructure for recommended process adjustments can then be fed back using the process workflow/ API mechanisms for achieving the desired outcomes.