The ability to connect industrial assets and machines using networks existed for a long time before the Industrial Internet of Things started to gain popularity. But what we’ve seen in the last five years is the emergence, and actually the need, for advanced analytic solutions that deliver a next-generation level of integration. It’s the critical component, when applied to machine and process data, which will enable more intelligent manufacturing operations and the push toward transformational business outcomes.
Data analytics has already become one of the fastest growing and lucrative new professions in industry, and it will continue to be a key focus of attention as it looks to combine skills in mathematics, computer science and engineering in ways that address the converging needs of IT and OT technologies within manufacturing companies.
Looking at new analytics architectures
A new Industrial Internet Consortium white paper looks into analytics architectures, and how industrial companies can get started with designing analytics systems. The report concludes that the required architectures need be analyzed in the context of defining end-to-end IIoT systems that functionally decomposed into five functional domains. These are:
1. Control: sensing, communication, execution, motion, and actuation;
2. Operations: provisioning, management, monitoring, diagnostics and optimization;
3. Information: data fusion, transforming, persisting, modeling and analyzing;
4. Application: logic, rules, integration, human interface; and
5. Business: enterprise and human resources, customer relationships, assets, service lifecycle, billing and payment, work planning and scheduling.
A key technology trend that continues to be a focus for industrial machine control is the ability to identify and recognize machine operational patterns, and make predictions that can result in better decision making and increased performance. The paper identifies three major categories or types of analytics that are being applied in industry.
|Mapping analytics into an effective Industrial Internet reference architecture requires combining functional domains that unify control, operations, information, application and business/enterprise systems. Image source: Industrial Internet Consortium|
Descriptive analytics gain insight from historical or current data streams including status and usage monitoring, reporting, anomaly detection and diagnosis, etc. Predictive analytics identify expected behaviors or outcomes based on predictive modeling using statistical and machine learning techniques. Prescriptive analytics finds an optimal solution by determining what is likely to happen based on first principles, using models and predictive analytics incorporating causality related to design and execution decisions. An example is on-demand production from a solid geometric assembly model to find the optimal set of manufacturing processes to achieve the final product intent, taking into consideration all possible options and capabilities.
Analytics key to IIoT developments
One fundamental prerequisite of industrial analytics is the availability of machine and process data but the good news is that industrial machine control systems have a plethora of data readily available for analysis. Now we’re seeing that combining data science and subject matter expertise (know what information is important) is vital to producing the best results. Automation and control suppliers working in areas such as machine condition monitoring and artificial intelligence are keenly aware of the need for machine/process expertise.
A final step will be to communicate and present the industrial analytics results in a compelling and easily understandable formats. As analytics advances, more and more meaningful operational patterns will be detected, identified and reported as alerts, and automatically reported to operators. Machine operational efficiency can be monitored and optimized based on analytics results to improve manufacturing and operations while at the same time reducing the stress on operators.
The report also concludes that “analytics is no magic by itself—it requires a combination of obtaining the proper data at the proper time, applying the proper analytics algorithms and models which are guided by the necessary engineering domain knowledge from both the machine manufacturers, system integrators and the plant operators themselves.”
The report is available here.