Contributing authors: Jeff Kelly, David Floyer
Not all Big Data is created equal. Data associated with the Industrial Internet – that is, data created by industrial equipment such as wind turbines, jet engines, and MRI machines – holds more potential business value on a size-adjusted basis than other types of Big Data associated with the social Web, consumer Internet and other sources.
This is due to the nature of the industry sectors whose high-value assets create the majority of Industrial Internet data. Industrial sectors such as healthcare, energy, and transportation impact outcomes across geographic, socio-economic and commercial boundaries.
A high obesity rate, for example, directly impacts not just the health of individuals, but also regional and even national economic output. Power outages, to take another example, can cost affected areas hundreds of millions of dollars in lost productivity.
On the flip side, improved operations, better efficiency and – most importantly – net new business models in the industry sectors whose daily operations impact each and every one of us, enabled by the Industrial Internet and Big Data analytics, can lead to improved economic and societal outcomes.
Harnessing the potential value of the Industrial Internet, however, presents a number of significant challenges not faced by those looking to tap into Big Data analytics in other, more consumer-focused industries. This is due in large part to the mission-critical nature of the workloads involved in the sectors that make up the foundation of the Industrial Internet.
To illustrate two extreme but useful comparisons, consider the impact of administering the wrong medication to a critically ill patient versus serving up the wrong display ad to someone browsing the Internet. Or the consequences of engine failure on an airliner 30,000 feet over the Pacific Ocean versus an online shopping application crashing during checkout. In both cases, the potential value and consequences of the former dwarf those of the latter.
The tools and technologies used to process, store, analyze, and optimize Industrial Internet Big Data, therefore, require significantly higher levels of performance, reliability, flexibility, security, and analytic complexity than their consumer Internet counterparts.
While Wikibon believes that Big Data is the new definitive source of competitive advantage across all vertical markets, the stakes are highest – financial and societal -- in those industry sectors that make up the Industrial Internet. The high value and consequences of harnessing this data, therefore, demand the development of specialized platforms, data models, and analytic capabilities to meet these unique, stringent, and critical requirements.
Opportunities by Industry Sector
Leveraging the Industrial Internet and the torrent of data being created by industrial equipment presents numerous opportunities to improve efficiencies of current operations in the impacted industry sectors. More valuable are the new business models and methods for delivering services that can be developed through the Industrial Internet and Big Data analytics.
Th following is an analysis of the opportunities available in three industry sectors: healthcare, energy, and transportation.
Healthcare spending in the U.S. topped $2.75 trillion in 2012, or 17.5% of GDP. That’s the highest spend on healthcare in the world(1). Healthcare analysts estimate that as much as 43% of that spending went towards unnecessary procedures and administrative waste(2). Meanwhile, healthcare outcomes in the U.S. actually lag that of other developed countries. The opportunities to leverage the Industrial Internet to improve efficiencies, eliminate waste, and improve patient outcomes in healthcare are many.
From a data perspective, high-value medical equipment, such as MRI and CT scan machines, create numerous data points through the course of normal operations. Lower-value assets, such as hospital bedside monitors and even patients, themselves, are increasingly being equipped with data-generating sensor technology, creating yet more data. The adoption of electronic medical records is also increasing, due in part to requirements and incentives included in the Affordable Care Act. Each medical record has multiple structured and unstructured data points associated with it.
All of these factors mean the volume and velocity of healthcare-related data are on a steep upward trajectory. Wikibon has identified a number of near-term application areas for the Industrial Internet in healthcare:
Maximize High-Value Asset Lifespan and Performance. Data generated by high-value assets such as MRI machines can be monitored and analyzed to predict the likelihood of part failure in advance to facilitate preventative maintenance. Taking preventative action to maintain high-value assets is orders-of-magnitude less expensive than repairing or replacing assets after failure has occurred. By better understanding high-value asset lifespans, hospitals and clinics can also plan better replacement and upgrade strategies. Using predictive analytics to determine likely regional illness patterns, which in turn impact the type and quantity of high-value assets required to treat them, can optimize this process further.
Improve Capacity and Workflow Management: Better understanding likely patient traffic patterns can allow hospitals to better allocate resources and staff, thereby reducing patient wait times and increasing patient satisfaction. For example, knowing when spikes in emergency room visits are likely to occur and the types of conditions likely to be presented based on analysis of historical patterns allows administrators to optimize staffing levels, equipment availability and room availability.
Ensure Process Compliance: Sensor technology embedded in low-value assets can have an equally significant impact on outcomes. Consider that 5% and 10% of hospital patients acquire new infections after being admitted, in many cases due to a lack of clinician compliance with hand washing guidelines(3). Sensor technology in soap and disinfectant dispensers can help ensure that doctors and nurses properly clean their hands when required (i.e., in-between visiting different patients and before entering sterile rooms) and significantly reduce MRSA infections among patients, which costs hospitals approximately $30,000 per case(4).
The long-term potential to remake the healthcare industry through better use of industrial and sensor data is even more dramatic. Specifically, by putting more data and insights into the hands of patients and by instrumenting patients themselves with data-generating sensors, clinicians and patients can change service delivery models.
Better-informed patients, for example, can make better decisions about when to visit the doctor in the first place. Tests can be scheduled and resulting data analyzed before patients visit clinicians, potentially eliminating unneeded hospital visits. Clinicians can remotely monitor and analyze data streaming off sensors affixed to patients to better ensure treatments are being followed and to offer proactive interventions, reducing costly hospital readmissions.
Based on Wikibon’s estimates, a 20% improvement in efficiency driven by the better use of Industrial Internet data could result in $2.1 billion in reduced spending on healthcare per year while improving outcomes for patients. The effects of the resulting health benefits across populations would likely ripple through the rest of the economy, leading to improved worker productivity and overall quality-of-life.
The energy sector is a complex organism, with multiple actors playing different parts throughout the energy lifecycle – from the discovery/creation, extraction and refining of natural resources through to the delivery of power (itself in various forms – oil, natural gas, electricity, etc.) to homes and businesses. At each step along the way, large and growing data volumes are created by energy-related industrial equipment that can be leveraged to improve efficiencies and ultimately develop new services for energy consumers.
Big Data-generating industrial equipment includes the machines and related software used to find and extract crude oil and other fossil fuels, both on land and at sea; power generating machines such as nuclear power generators and wind turbines that feed the energy grid; and smart meters that control and monitor power consumption at the point of delivery.
According to a study by the University of Minnesota(5), power outages and “power quality disturbances” cost the United States as much as $188 billion per year. A 13-hour power outage in the San Diego metropolitan area in September 2011 resulted in an economic cost of close to $120 million in the form of perishable food losses, overtime pay to government workers to deal with the fallout, and lost worker productivity(6).
Less dramatic but equally important are the everyday efficiencies – or lack there of – of power delivery and consumption. The opportunities to improve efficiencies in the energy sector are plentiful. To illustrate the potential of Big Data analysis in the energy sector, Wikibon investigated possible applications at two points in the energy lifecycle - energy creation and energy consumption.
At the energy creation end of the spectrum, consider large-scale industrial wind farms. These generators are made up of hundreds of individual wind turbines. Wind turbines are highly mechanical machines, with hundreds of moving parts such as blades, gearboxes, and converters. They also contain hundreds of data-generating sensors measuring things like wind speed, pitch and yawn degrees, oil temperature, etc.
Such data, which is generated as frequently as every few milliseconds, can be used to better understand wear-and-tear of the various parts that make up a single turbine. Wind farm operators can use this data to predict when part failure is imminent and take preventative action. Taken a step further, sophisticated analytics as applied to aggregate wind turbine data can allow operators to optimize operations of an entire wind farm, ensuring that individual turbines are working in concert to meet specific output objectives.
By conservative estimates, Wikibon believes use of machine-generated data streaming off industrial wind turbines to facilitate preventative maintenance would extend the lifespan of the average wind turbine by three years, increasing total average lifespan to 18 years. Such an increase would result in a 17% reduction in costs per year per wind turbine.
On the energy consumption side of the equation, smart meter technology that tracks and monitors energy usage data every 15 minutes is increasingly replacing aging, traditional meters that require manual readings. The result is a deluge of highly granular machine-generated usage data. With such data at its disposal, utility companies are in possession of the raw material needed to dramatically improve efficiencies, reduce downtime, and improve customer service.
Specifically, intelligent use of smart meter data allow utilities companies to:
- Better monitor and forecast energy consumption patterns;
- Identify inefficient energy use at both the macro and household levels;
- Accurately predict potential power outages and equipment failures;
- Improve customer segmentation and tailor service offerings based on customer behavior.
Utilities companies that take such approaches will both reduce the amount energy and money wasted due to inefficiencies and potentially identify new ways to package and productize energy deliverables to increase revenue. Instead of reacting to changing market conditions and consumption behavior, utility companies can be proactive in their approach to energy efficiency and management.
The average U.S. home consumes 11,280 kWh of electricity per year. Not taking into account the better use of electricity via new analytic-driven services, smart meter technology is likely to reduce electricity consumption by 3%-5% per year per household.
Ultimately, Big Data analytics coupled with Industrial Internet data could allow for the optimization of processes across the entire energy sector – from data creation to data consumption – vastly improving the efficiencies and reducing wasted resources.
Like the energy and healthcare sectors, the transportation sector is ripe for innovation in the form of Industrial Internet and Big Data analytics. This sector includes the locomotive, trucking, and commercial airline industries, among others. For this research, Wikibon examined the potential impact of the Industrial Internet and Big Data analytics on the commercial airline industry.
Jet engines create significant data exhaust during operation. The new Boeing 787, for example, creates as much as a terabyte of data per round-trip(7). This data can and should be analyzed to increase operational efficiencies and facilitate preventive maintenance of faulty and soon-to-fail parts. Big Data analytics can also play a critical role in the design and testing of new jet engines and other aircraft equipment.
The result of Big Data analysis as applied to the commercial airline industry should result in benefits for the airlines, their passengers, suppliers, and the environment. For example:
Preventative Maintenance and Improved Design: By analyzing data created by jet engines and sensors monitoring the surrounding environment (temperature, humidity, air pressure, etc.), service providers can predict when various parts are likely to fail and take preventative maintenance actions. As in healthcare scenarios, replacing a soon-to-fail jet engine part before it malfunctions is significantly less costly – and safer for all involved - than doing so after the part fails during operations.
Fleet Management and Schedule Optimization: Preventative maintenance reduces aircraft “down time,” meaning a higher percentage of the fleet is available to service passengers. By further leveraging aggregate data associated with arrivals/departures, weather conditions, and other data sources, airlines can better manage their fleets and reduce the number of delayed and cancelled flights.
Develop New Services and Routes: Improved customer satisfaction is the likely result of fewer delays and cancellations, potentially increasing customer loyalty and ultimately increasing bookings. By analyzing customer flying patterns, airlines can also identify profitable new routes to add and other services that will benefit both customers and the airline’s bottom line. More efficient jet engines also consume less fuel and emit fewer environmentally contaminating gasses.
More exciting, the Industrial Internet also provides opportunities to reshape the relationship between jet engine manufacturers and the airlines. Rather than pricing per engine, manufacturers could charge airlines based on their use of each engine as defined by operational data. This arrangement would encourage airlines to make more efficient use of and better maintain engines, while providing manufacturers a steady stream of data with which to design next generation engines.
Similar benefits can be achieved in other sub-segments of the transportation sector through the application of Big Data analytics enabled by the Industrial Internet. The operations of train engines, just like jet engines, when instrumented with data-generating sensors, can be optimized to avoid costly breakdowns and improve design. Shipping companies that outfit truck fleets with sensor technology can leverage the data generated to identify more efficient routes and improve fuel efficiency.
The ultimate goals across these three sectors – healthcare, energy and transportation – as well as other industrial sectors like manufacturing are to leverage the Industrial Internet and Big Data analytics to improve operational effectiveness and identify new ways of serving customers. By orchestrating and optimizing processes across operational silos and across sectors, and by using advanced analytics to identify and exploit valuable new business models, the Industrial Internet offers these and related industries the opportunity to truly revolutionize themselves for the good of all involved.
Challenges and Platform Requirements
To truly leverage the Industrial Internet across these and other industry sectors, new platforms, data models and analytic capabilities are required. Wikibon has outlined the stringent requirements such technologies must meet in Defining and Sizing the Industrial Internet. These requirements are summarize below, with illustrations taken from the healthcare, energy and transportations industry sectors.
Data volumes associated with the Industrial Internet are growing at twice the pace of other sources of Big Data, including social media, according to Wikibon’s analysis. A single wind farm made up of 500 turbines creates as much as 2 petabytes of data per year, for example. And each new generation of industrial equipment contains more sensors and creates ever more data. Any platform used to collect, store and analyze Industrial Internet Big Data must, therefore, be linearly scalable and take advantage of open solutions such as Hadoop.
As mentioned, even a relatively brief power outage can cost the effected area millions of dollars in lost productivity and other costs. In healthcare scenarios, the loss of operational capacity is literally a matter of life or death. And the consequences of a jet engine failure in mid-flight are obvious. Clearly, platforms to support the Industrial Internet must be highly available and contain multiple failover and immediate recovery capabilities.
Cyber-warfare and cyber-crime is on the increase. Both state and non-state actors are increasingly attacking infrastructure targets such as energy grids and national defense assets via the Internet. In healthcare scenarios, personal medical information is covered by numerous security and privacy regulations. Industrial Internet platforms therefore must be highly secure and be constantly reinforced with new methods to keep up with developing threats.
Flexibility and Openness
The Industrial Internet is made up of too many technologies, software, and machines to count. The results are numerous data types, numerous workload requirements, and numerous analytic requirements. Some analytic processes, for example, must occur on edge devices, while other analytic workloads are better run in centralized or cloud environments. In some industries, regulations stipulate that certain data must reside inside corporate firewalls and not in public cloud environments. Any platform used to leverage the Industrial Internet must be flexible to accommodate this heterogeneous and ever-changing environment of technologies and open standards. Further, where possible industry-wide standards should be developed and adhered to.
While initial Industrial Internet applications will likely focus on single applications, eventually companies in industrial sectors will want to orchestrate multiple applications to work intelligently together with the goal of optimizing their entire operational environments. The results of one analytic process, for instance, will kick-off a series of corresponding actions, which lead to further analytics and yet more consequential actions. It is critical that a highly networked platform to support this type of environment give companies the ability to model and define these workflows and incorporate machine learning to optimize analytic decision-making in real-time.
Another inhibitor to the development of the Industrial Internet is creation of the data itself. While it is true that more and more industrial machines are being outfitted with data-generating sensors, many white spaces still need to be filled. Jet engine makers can’t simply slap new sensors on engines wherever they please. As small and light as modern sensors are, they still impact the physical operations of equipment, and their implementation must be accounted for in the design process.
Beyond technology requirements, policy and legal issues must also be addressed to fully leverage the Industrial Internet. Depending on industry, certain data types cannot be shared with certain parties. In the commercial airline industry, for example, rules prohibit certain flight data being shared with the airlines. In the case of nuclear power plants, most data sources cannot be connected to the public Internet by law. These and other issues that hamper data sharing, and access must be addressed.
Action Item: The opportunities to improve efficiencies and create valuable new business models associated with the Industrial Internet are vast. To do so, however, requires the development of specialized platforms, data models and analytic capabilities to meet the many unique and critical requirements associated with industrial data, workloads, and processes. The companies that make up the industrial sectors, manufacturers of industrial equipment and technology vendors must work together to develop the platforms and technologies needed to leverage the Industrial Internet, with particular attention paid to the opportunities and requirements outlined above.
Footnotes: (1) List of countries by total health expenditure (PPP) per capita, Wikipedia http://en.wikipedia.org/wiki/List_of_countries_by_total_health_expenditure_(PPP)_per_capita
(2) Reducing Waste in Healthcare, Health Policy Briefs http://www.healthaffairs.org/healthpolicybriefs/brief.php?brief_id=82
(3) Hospital Epidemiology and Infection Control in Acute-Care Settings, http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3021207/
(4) Targeted versus Universal Decolonization to Prevent ICU Infection, http://www.nejm.org/doi/full/10.1056/NEJMoa1207290
(5) The Rising Tide of Power Outages and the Need for a Stronger and Smarter Grid http://tli.umn.edu/blog/security-technology/the-rising-tide-of-power-outages-and-the-need-for-a-smart-grid/
(6) Economic impact of September 9 Power Outage: Conservatively estimated at $97 to $118 Million http://www.nusinstitute.org/assets/resources/pageResources/PrelimReportSDBlackoutEconImpact.pdf
(7) 787 Dreamliner, Techinical Information, Boeing http://www.boeing.com/boeing/commercial/787family/specs.page?
For a list of Wikibon clients, click here.