BI Dashboards Integrate Smart Factory Data for Meaningful Analytics
The advent of industrial IoT and smart factories has put new requirements on next-generation analytics systems to unlock operations data in novel ways. While smart industrial analytics is a relatively recent use-case, activity is beginning to heat up. The challenge will be finding meaningful trends from the data extracted from multiple industrial IoT touchpoints, beyond simply storing the content in operational logs.
Smart factory generated by Internet of Things (IoT) sensors must be correlated with other corporate data points, and the search for meaning must become a regular part of everyday workflows, not a fleeting moment.
Integrating IoT data into regular processes involves requires analytics software. This software is fueled by artificial intelligence AI and machine learning technologies. Another part of the equation is IoT connectors that link IT business intelligence dashboards with factory operations data.
While BI dashboards have long been a part of backroom analytics, most haven’t been able to adequately process industrial IoT data inputs until recently. For smart factories to avoid siloed data, it’s paramount to select BI dashboards equipped with capable analytics. Today many dashboards combine industriaI IoT with access to data lakes – vast storage pools intended to aggregate large amounts of unstructured information – or else cloud databases.
“Smart factory data has much in common with the data that comes from other functions across a business,” said Enno de Boer, partner, McKinsey. “To be of value, it must be used to inform decision making,” Otherwise, there is little point in harvesting and aggregating vast quantities of data.
Across the Value Chain
To be truly valuable, factory floor data must be integrated throughout the entire value chain, said de Boer, who heads McKinsey’s work in digital manufacturing and its collaboration with the World Economic Forum as part of the Global Light House network.
With better use of analytics, de Boer sees tailored production that influences everything “from component sourcing through to last-mile delivery.”
Business intelligence analytics today is a common feature of enterprise IT products. But applying the technology for operations has proved more difficult. Despite blockers in implementation, the global is expected to reach $16 billion by 2026, according to ResearchAndMarkets.com.
Smart Factory Analytics Scorecard
Multiple vendors now strive to deliver improved industrial analytics and BI dashboards. Players in the forefront of the smart factory market include ABB, Honeywell International, Robert Bosch, Siemens and others.
When it comes to capturing, processing, storing and analyzing smart factory data, IT giants with notable footprints in manufacturing are part of the mix. Chief among these are IBM, Hewlett Packard Enterprise and SAP. Innovative data startups have also targeted the specialized requirements of smart factory analytics, such as Cloudera and DataStax.
As cloud becomes the locus point of factory analytics, cloud leaders Amazon Web Services, Google and Microsoft are building specialized data workflow pipelines. Players in turn support end-user business intelligence dashboards such as those from Looker, Microsoft, Tableau, and others.
Smart Factory Buildout
Building out smart factory analytics is a formidable task. A typical manufacturing site can create more than 2,200 of data in a single month, and most of that data is unanalyzed, according to an IBM report on digital transformation. The influx of data that remains unanalyzed contributes to the problem of industrial IoT proof-of-concept (POC) projects that drag on.
Most industrial data is generated outside of IT, emphasizes Manish Chawla, general manager for industries, energy, resources and manufacturing at IBM. He indicated that recent industry efforts focus on improving project foundations; poor planning can prolong the lead time of POCs.
“People tried to build a penthouse without having a foundation,” he said.
Chawla also said IBM had recently worked alongside Siemens and Red Hat on a cross-platform approach to execute analytics from Siemens’ Industrial IoT platform, MindSphere, closer to the factory edge.
SAP is working to allow customers to analyze a mix of time-series-oriented historian data along with IoT and business data, said Dominik Metzger, VP and head of product management, manufacturing and Industrial IoT, SAP. A data historian is a software function that logs the output of manufacturing IT processes for governance purposes.
For Metzger, one of the key changes in recent years is the degree of standardization in data handling. “It’s gotten more economical, and scalable,” Metzger said, citing data lakes as an enabler of analytics for smart factories.
SAP views embedding IoT data analytics in business processes as a key next step in its Industry 4.0 strategy, which it has dubbed Industry 4. Industry 4 is a reference architecture that spans workflows from sources such as data historians, edge services, and cloud or ERP systems with business intelligence capabilities.
Analytics Require Data Volume
The evolution of smart factory analytics is complicated by forces affecting analytics generally. For instance, the rise of predictive and prescriptive analytics based on AI and machine learning presents several implementation challenges. Here users should proceed thoughtfully when using analytics to delve deeper into operations, according to Ed Cuoco, vice president of AI and Analytics at PTC.
When implementing analytics for diagnostics, for example, there are times when simple statistical process control might be preferred over machine learning or AI-type solutions, Cuoco said.
“Without good-quality historical data in volume, you may not be able to derive the insight you want,” he added.
IoT platform provider PTC works closely with end users and other software makers to serve up analytics from the factory front line to the business end user, and sometimes back again. That’s the case with a recent deal that sees the Fujitsu Smart Factory framework using PTC’s Vuforia augmented reality and ThingWorx platforms to convey analytics information to operations workers.
Novel Graphics for Analytics
Graph data technology — long on the periphery of the advanced data analytics scene – has gained acceptance in factories and other settings. Graph databases such as Aura Enterprise from Neo4j have proved useful and put users’ smart factory analytics into context and enabled collaborative projects that identify new operational efficiencies.
Unlike relational databases that underpin the bulk of data analytics and store data in rows and columns, graph data formats use data mappings to manage complicated connections between data elements. Neo4j’s target sectors include automotive, warranty, analytics, supply chain management and medical instruments. The medical sector in particular had demonstrated graph databases’ ability to foster cross-team collaboration, according to Amy Hodler, director of graph analytics and AI programs, Neo4j.
A medical instrument company looking to track failures before product shipment found Neo4j’s graph methods useful, Hodler noted. Identifying such failures generally involves detective work because all subcomponents of a faulty instrument must be traced to determine whether they are responsible for the failure.
To put analysis in the hands of more users, Neo4j offers connectors that link its graph data models to data visualization and discovery dashboards such as Tableau, Tibco Spotfire, and others. The company offers its own Bloom visualization tooling as well.
Also connecting to a host of visual dashboards are software management tools from DataStax, a company that largely led to the commercialization of the open source NoSQL database. The enterprise edition of DataStax’s product supports graph data handling. Among the IoT application creators using its software is South Africa-based Locstat, which deployed the product to analyze sensor data and real-time streaming analytics.
“Visualization is becoming an increasingly important element of trying to understand what is happening in the IoT landscape, in particular when you’re dealing with a fairly complex setup,” said Matthias Broecheler, chief technologist at DataStax.
The visual analytics tools help operations staff, developers and others, he added. At the same time, Broecheler noted that some decisions in smart factories require immediate response. That driving force is behind new forms of analytics processing that, without human transformation, autonomously detect and respond to factory floor anomalies.
Goodbye, Data Silos
In smart factories, managers, field operations and IT development teams need to work together just as in any other kind of business transformation, said McKinsey’s de Boer.
“Transformations fail when teams operate in silos, and only one function drives the attempts to initiate changes,” he said in an email interview. The drive to democratize data requires people across the entire organization to understand the power of new technologies, and how to use them, de Boer said.
For the manufacturing sector, the role of operations personnel in determining data democratization will be telling.
“With analytics tools in the hands of operations personnel, companies will be able to more easily develop solutions that answer business challenges,” de Boer said.
McKinsey’s de Boer pointed to the analytics academy programs set up by members of the Global Lighthouse Network and argued that all stakeholders could gain from participating, including everyone from the boardroom to production frontlines.