Is Data The New Oil? It’s More Complicated

Explore the value of data to an industrial plant and the importance of Automated Machine Learning with the context of a plant’s digitalization processes.

It’s been said that data is the new oil of the 21st century. That is an over-simplification. Oil is not a renewable resource, whereas data is in abundance and continuing to grow. Data growth is dramatic, in fact 90 percent of world data has been generated over the last two years. At the same time, in their natural form both lack value. For crude oil, the value is derived only after it has been distilled and refined. For industrial data to provide valuable insights, it needs to be cleansed, processed and analyzed using Machine Learning and AI technologies. Oil fuels the industrial factory, whereas information fuels and controls the digital factory.

Let's explore the value of data to an industrial plant and the importance of Automated Machine Learning with the context of a plant’s digitalization processes.

Industry 3.0 and the Application of Sensor Data

Let’s start with how data is used today. A typical industrial plant has anywhere between thousands to tens of thousands of sensors embedded in its machinery generating terabytes of data. These sensors monitor the physical properties of the machine including temperature, voltage, vibration, pressure and flow. Sensor signal data plays a critical role in the control process of industrial machinery, serving as inputs and checkpoints for the controller rules that govern the machine operation. Some data is displayed in real-time on screens in factory control rooms for human controllers to view and react to machine status.

Typically, some of the sensor data is stored for years on local servers for post-mortem analysis. When a machine fails critically, a process engineer may be called to analyze the failure data to understand the root cause and adjust control rules accordingly. In most cases, industrial plants ignore or forget their data. At best, data is used in a small number of scenarios, but is never fully operationalized.

Some plants view data the way they do non-hazardous industrial waste. They are byproducts of production processes and need to be captured, stored and ultimately disposed of. Although both can be monetized, they are traditionally viewed as cost areas.

Data in the IIoT Era

With IIoT, both the economics of Big Data and the science of Machine Learning have come together.

A confluence of multiple factors has lowered the cost of data transportation, bandwidth, storage and analysis. For example, data storage has fallen from $569 per GB in the early 1990’s to less than $0.01 today.  

It was not long ago that Artificial Intelligence was an esoteric concept studied within institutions of higher learning. In recent years, the potential value from the application of Machine Learning to the industrial domain has been recognized. Significant investments in the domain have been channeled to both startups and existing solution providers’ R&D.

Now with the cloud revolution, every industrial plant is connected with high-speed and low-cost information pipes to the cloud, where data can be processed with unlimited processing power at minimum cost.

Automated Machine Learning for Real Time Predictive Maintenance

At Presenso, we believe that hidden within the terabytes of sensor-generated data are micropatterns that can be used to detect abnormal machine behavior. The abnormal behavior is used to alert production plants of evolving asset degradation or asset failure. The key to extracting insights from the large data heap is automation. When software can now automatically extract insights, without human input or intervention, the full potential of a scalable solution can be achieved.  

So why are we not seeing the full potential of data?

Until now, the existing statistical packages have lacked the power to analyze data in real time. Existing solutions are highly depending on human expertise, and there is a severe shortage of qualified Big Data scientists and engineers within the industrial domain.

Data may be the oil, but it requires an accelerant to become operational. The accelerant is Automated Machine Learning.

With Automated Machine Learning, advanced algorithms are applied to big data in real time and extract operational insights. To scale beyond a small number of machines, the correct algorithm needs to be applied without the human input. Since the Machine Learning is applied in a cloud environment, all the data generated form a single facility or multiple facilities can be analyzed in real time.

Follow the Money: Financial Opportunity from Big Data

Applying Automated Machine Learning to Big Data can help drive top line revenue growth, reduce manufacturing costs meaningfully and improve the balance sheets of industrial facilities.

It is estimated that $650 billion is lost to industrial machinery downtime per year which represents 5 percent of the $13 trillion in global manufacturing output. Downtime is expensive because of the opportunity cost of lost production. Recapturing the 5 percent of lost production improves plant yield rates. At a global level, it can impact both economic development and the environment.

When machinery fails unexpectedly, it creates a crisis within the plant as technicians search for the underlying cause. Repair crews may need to be re-assigned from existing job sites and new parts are ordered. With IIoT Predictive Asset Maintenance, industrial plants can significantly reduce downtime. That is because plant technicians are alerted of evolving failure before it occurs, giving them time to reduce production workloads, order parts and schedule maintenance before production stops.

There is also a balance sheet impact. When Big Data will be used to determine the timing of frequency of repairs, it will likely improve the lifetime economic value of industrial machinery, resulting in lowered Capital Expenditures. Furthermore, today industrial plants are forced to overinvest in redundant machinery and inventory to protect themselves against unscheduled asset failure. As the incidence of unscheduled asset failure declines, investments in redundancies can be reduces, thereby improving the health of plants’ balance sheets.

Summary and Conclusion

As industrial plants move to digitalize their production assets, Automated Machine Learning is the key to unlocking the value of sensor-generated data. Whether or not data is the new oil is not important unless it is coupled with Automated Machine Learning. The accelerator that can transform data of limited value to the driving force of Industry 4.0 and the Smart Factory.

David Almagor is Chairman of Presenso.

More in Industry 4.0