As we know, the Internet of Things generates huge quantities of usage and sensor data. However, companies only use such data – in logistics, maintenance, or other areas – if they can analyze it extremely quickly. This mass information can be found in both structured (such as a database table) and unstructured formats (a blog, log file, or e-mail, for instance). By efficiently merging all of these different data types and applying corresponding algorithms from the realms of statistics and business intelligence, it’s possible to glean valuable information. A good deal of experience and tangible data from past situations can indicate when a certain mechanical component should be replaced before a breakdown occurs.
Combined with asset management and other solutions, these and other predictions can be applied to existing installations and products that are currently being manufactured or already being used by customers. “A quality assurance system that supports supply chain planning, root cause analysis, quality notifications, and complex product-quality planning can reduce scrap by 40 percent,” states Scott Bolick in a new SAP whitepaper on performance benchmarking. A quality assurance system of this kind – including predictive maintenance – is already in place at the German automotive company Audi AG.
Frequency analysis can help predict when a tool will break down
“The self-maintaining functions of lathes and other production machines are going to be an integral part of standard maintenance,” explains Rüdiger Spies, an analyst at the market research company Pierre Audoin Consultants (PAC). “These modern lathes take measurements on their processing runs even as they work through them.” According to Spies, frequency analysis then makes it possible to determine whether a given tool is about to break down. “Along with replacing the tool, it’s necessary to account for this change in optimizing your production planning, as well.”
With statistical methods like multivariate outlier detection, it’s possible to identify anomalies early on, such as in a power plant’s turbine operations or a retailer’s supply chain. The energy company Israel Electric, for instance, managed to increase its pre-warn time from 30 minutes to 30 hours, and a light alloy foundry reduced its scrap by 80-90%. When they automate these predictive and analytical methods and integrate their corresponding processes, companies can achieve a level of responsiveness you can measure in milliseconds. “A standardized data model, fully integrated processes and equipment, and analytical tools for predictive maintenance give you the three-part harmony you need to ensure further growth and a sound, transparent business model,” explains Ralf Thiemann, senior managing consultant for energy and utilities at IBM Global Business Services.
Rüdiger Spies and other experts believe that the “fully integrated production of the future” – also known as Industry 4.0 – will give rise to further forms of predictive maintenance. Automotive parts like engines, gear mechanisms, and clutches, for example, will be capable of self-diagnosis and other intelligent functions. This will give employees in production and maintenance new ways to take advantage of early-warning systems and prevent breakdowns.
Service and maintenance cost nearly 30% lower
Ultimately, integration and analysis are the keys to succeeding with predictive maintenance. “The companies surveyed for SAP’s ‘Idea to Performance’ whitepaper reported meeting their production plans 19.7% more often when they coordinate customers’ delivery dates with real-time manufacturing conditions and material availability,” offers Scott Bolick. “Meanwhile, organizations that gather and analyze data at all of their locations for the purpose of standardized plant management spend 29% less on service and maintenance services.”
In Industry 4.0, business IT needs to dovetail with the Internet of Things. The required components are already available: sensors that monitor the performance and status of machines; RFID; mobile network connectivity (mSIM); and powerful software capable of interpreting data, developing recommendations, and automatically carrying out tasks. To achieve predictive maintenance, companies have to integrate maintenance logs, configuration data, and sensor and telemetry information from cyber-physical and other systems. All this can be handled in SAP Predictive Analysis. Based on the SAP HANA platform, this software’s predictive analysis library can be combined with the statistical techniques provided by the R environment and information from a Hadoop database cluster. SAP HANA then performs the necessary calculations very quickly and presents the results in SAP Predictive Analysis.