The Ultimate Survival Guide on AI and Machine Learning in the Fourth Industrial Revolution, Part IV

In this five-part series on industrial AI, we will arm you with the knowledge you need to understand what AI, machine learning, data science and edge computing are, and how they are already impacting industry.

See part three, where we covered how data science fits in and why data scientists are so important.

PART IV: WHAT ARE THE PRACTICAL APPLICATIONS OF MACHINE LEARNING IN INDUSTRY?

The impact of incorporating new technologies is huge. Across eight industrial sectors, the value at stake is upward of $17 trillion.

When machine learning is applied to industrial data, it opens up new doors for today’s businesses to improve their operations and save time and money. They can predict and prevent equipment failures before they happen, improve the availability and reliability of their critical assets, and cost-optimize their maintenance programs. This comprehensive approach is known as asset performance management (APM).

With APM, industrial businesses can make critical decisions about their assets based on known failure modes, monitor machines to determine which require maintenance based on historical patterns and real-time conditions, and make critical future decisions about assets based on predictive insight.

Anomalies that deviate from normal behavior patterns can be detected, relevant signals that would otherwise be buried in mountains of data can be extracted from less-relevant noise, and machine failures can be predicted and prevented before they occur.

Understand Everything About Your Equipment with Asset Performance Management

APM is fueled by three main technology areas that can help businesses make faster, smarter decisions that are based on data instead of intuition:

  • Preventive Maintenance uses expert historical knowledge about assets — such as equipment types, failure modes and maintenance tasks — to inform more cost-effective maintenance strategies and help minimize risks and failures.
  • Predictive Analytics uses data and machine learning to anticipate what might happen in the future. Businesses can use historical, real-time and contextual data about the environment to make predictions about what is likely to transpire next — surfacing overlooked problems and opportunities.
  • Prescriptive Analytics uses data and machine learning to provide data-backed recommendations on the right action to take to achieve specific business goals. This typically involves human interaction (also known as Human-in-the-Loop) to validate the recommendations and take action. This means the knowledge base automatically gets smarter — and its recommendations grow increasingly accurate — over time.

The value of APM in industry is enormous because of the power and potential of these technologies to unlock insights from data. For example, when it comes to mission-critical assets like airplanes, locomotives, bulldozers, trucks and wind turbines, the ability to associate a certain engine vibration in those machines with a pending component failure enables more effective repairs, greater safety, increased uptime and better customer service. Those assets can be sent to the shop for proactive service repairs before they break down mid-mission.

Increasing Business Productivity

Companies can use data-driven insights to improve APM and ensure their machines consistently hit peak performance.

In the energy industry, wind power development has grown dramatically in recent years. But as renewable procurement auctions press for ever-lower costs to find energy buyers, the industry can and will have to produce more energy from installed wind turbines to maximize investments and prevent underperforming preconstruction energy assessments. The industry can conquer this challenge by improving asset availability and producing more energy from existing turbines — with no new hardware required.

The data that is already captured by today’s installed wind equipment has the potential to transform many areas. However, the most immediate and profitable opportunity is in operations. Downtime in the current U.S. wind fleet prevents significant renewable energy from being generated — enough to power 1.1 million homes every year. And it prevents us from removing 8.9 million tons of CO2 out of the atmosphere annually.

Using data, problems can be identified before they happen to optimize operations and eliminate downtime. For example, within just a few days of deploying machine learning software, Berkshire Hathaway Energy was able to identify a predictive maintenance need on a wind turbine with a faulty bearing, saving $250,000 in equipment damages and days of costly downtime had the issue gone unnoticed.

Improving the Reliability of Critical Assets

By leveraging machine learning, companies can digitize their workforces and improve key operational workflows.

In the transportation sector, railroads have the opportunity to gain an inside edge against industry competitors and other modes of transit. They now have a viable way to increase the productivity and reliability of trains, significantly reduce fuel costs and environmental impact, predict and prevent locomotive failures before they happen, and improve the safety of their operations.

Flagging issues like oil pressure and engine failures used to mean operation analysts had to manually diagnose thousands of locomotive asset faults on a daily basis. This is a costly and ineffective process. With machine learning, potential problems get flagged in advance so that repairs can be made before locomotives even leave the railyard — or, if necessary, locomotives can be pulled off of the track at the optimal time for servicing.

Today, advanced railroads are using data to run virtual stress tests on locomotives to assess their health — eliminating the need to burn 250 gallons of diesel fuel per test while still providing the information railroads need to run safely and efficiently. Doing this across all the locomotives in the U.S. would save 17.5 million gallons of diesel fuel annually, enough to power 3,500 diesel locomotives for one year.

Making Industry Safer and More Secure

Industrial environments are becoming connected faster than they can be secured. By 2020, 25 percent of all identified cyber attacks in enterprises will involve the IoT. While the majority of executives have finally learned to understand the threats targeting their information technology (IT), much work remains to help them understand and address the severity of the emerging vulnerabilities and threats to their industrial operational technology (OT) environments.

Cyber attacks do not have to be inevitable. Companies can take proactive steps to minimize risks to their operations while increasing the safety, productivity and reliability of their digital and physical environments.

Data science and machine learning can be used to quickly identify threats and recommend proactive, preventive actions. These technologies help businesses become more cyber secure by:

  1. Acquiring and ingesting OT network, enterprise IT, operational and contextual data from disparate sources across the enterprise.
  2. Cleaning and normalizing data for processing and analysis.
  3. Applying machine learning models to identify bad actors and potential threats, and provide alerts and recommendations for remediation.
  4. Providing workflows for engineering and security operation center analysts to triage, investigate, remediate and resolve potential cybersecurity threats.

Improving the security of critical infrastructure helps protect a business’ most valuable asset: its people. The power of data science and machine learning enables today’s companies to address highly complex OT security issues before they spread throughout the organization.

In part five we’ll discuss the real-world applications of AI and machine learning in industry.

Don’t want to wait? Download A Survival Guide to AI and Machine Learning in the Fourth Industrial Revolution today.

GET THE SURVIVAL GUIDE NOW