How do we address the challenges in integrating robotics data with operational and enterprise systems?

Question

Accepted Answer

Adoption of industrial robotics has gained momentum in recent years, however companies are struggling to maximize their value through the massive amount of data that is created in real time. As a result, data integration is one of the most challenging problems with industrial robots to solve. The end goal should be the scaling of value from digital investments and unlocking the full potential of data.First, organizations must address the fundamentals before trying to harness the value of the data:An upgraded network infrastructure to handle the traffic from data proliferation, especially in remote plant sites, e.g., underground mining facilities, that have slow internet connection undergroundAn integrated ERP in place to monitor and manage the manufacturing process, transportation and inventoryAssuming you’ve already mastered “walking”, then it’s worthwhile to look into how to “run”.Most of the challenges around data integration across operational and enterprise systems are rooted from a misalignment of IT (Information Technology) and OT (Operational Technology) priorities and architecture.Such misalignment is primarily driven by different end-user expectations: IT teams are expected to make core IT operations efficient and stable at reduced costs, while OT teams concentrate on supporting real-time actions and rapid response to site-specific technology needs. This difference eventually leads to different IT and OT budgets / goals and incompatible architecture – making data integration between the two difficult to manage. Organizations must consider the following measures to affect the changes needed to effectively integrate data across industrial robotics, OT and IT systems:1. All data is not created equal. Focus on data that matters. Start with a closer look into the data demands from both operational and enterprise perspectives and define the integration plan accordingly. For example, the Force Torque Sensor (FT sensor) on robotic arms capture the force that the machine is applying with its end-of-arm tooling. The data captured by FT sensor has applications at the site level (e.g. assembly, hand-guiding, teaching and force limitation) and should be primarily stored and processed in the Operational Systems to lower latency / enable real-time analytics and avoid unnecessary pressure on network infrastructure. However, it doesn’t necessarily mean that the data should not be integrated into Enterprise Systems - a potential use case of the data at the enterprise level is to leverage advanced data analytics for predictive maintenance of these machines across all sites. While you may need real-time access to data at the site, frequent (e.g. monthly) update to enterprise system will support the analytical decision making on maintenance.2. Define data ownership and governance to ensure organizations can realize the value from data. A clearly defined data ownership and governance can ensure the accuracy and most efficient use of the data.Business stakeholders should be responsible for owning and managing the data while the IT team supports the storage and processing requirements. For example, a manufacturing equipment manager may be owning all the data captured by sensors on robotics for real-time monitoring while a plant manager may be responsible for the throughput and downtime of multiple production lines. In case of manufacturing equipment or production line upgrade, the equipment / plant manager needs to be responsible to refresh the data requirements (e.g. capture, storage, process, etc.) and use cases to reflect the change in data (e.g. more data from additional sensors). Before any digital transformation project, data ownerships and governance need to be defined to ensure the most appropriate use of data, which is often overlooked and prohibits organizations to fully leverage their data as a result. Data that is effectively integrated but not governed may become stale and ineffective for decision making.3. Leverage a data lake as the technology backbone to balance both structured and unstructured data. With massive amount of complex data being produced from robotics in every second, a data lake should be considered as the central repository of all forms of data, both structured and unstructured. Traditional data warehouse, optimized to analyze relational data coming from transactional systems and line of business applications, is not suited for the storage of data generated from robotics because it is captured by various sensors / cameras and has many forms (e.g. temperature, force, video, sound, movement, etc.). As all data is being stored in the data lake in a non-defined way without careful design or the need to know what questions you might need answers for in the future, it can enable organizations to run different types of analytics like SQL queries, big data analytics, full text search, real-time analytics, and machine learning to uncover insights. The use cases of unstructured data need to be defined, under different stages of excellence as the data mining tools / capabilities mature. However, the data lake is not a substitute for the traditional data warehouse as organizations still need data cleaned, enriched and transformed and it serves as single source of truth for enterprise data and reporting (e.g. financial performance). Organizations need to be equipped with the technology and capability to manage both structured and unstructured data.4. Define proactive security measures for both data at rest and data in transit for High Value Assets (HVAs). With the adoption of robotics and the manufacturing IoT sensors that comes with them, more data is created and shared across sites, across OT and IT systems, and across internal and external users. While developing your data integration plans, focus on identifying and protecting the data related to High Value Assets (HVAs, i.e. physical assets, business processes, and/or critical programs that are critical for your business operations). Proactively define security measures to identify at-risk data and implement effective data protection solutions (e.g. firewalls, auto-encryption, network access control) for both data in transit and at rest. While data in transit and data at rest may have slightly different risk profiles, the inherent risk hinges primarily on the sensitivity and value of your data – as you think about measures to identify and protect the data, don’t forget to invest in effective response and recovery measures.To learn more about the impact of digitization in industrial manufacturing, please visit our Digital Transformation webpage, our Manufacturing Center of Excellence webpage, our Future of Production webpage.