X

Learn about data lakes, machine learning & more innovations

Data in Action: IoT and the Smart Bearing

The Internet of Things (IoT) represents a big wave of technological change, and organizations in virtually every industry will benefit from this technology. Some 4.9 billion connected objects will be in use this year, up 30 percent from just last year, predicts research firm Gartner. By 2020, it adds, that number will increase to some 25 billion connected objects worldwide.

Businesses in many industries are evaluating the use of IoT technology for remote monitoring and to improve maintenance for mission critical operations. In a recent article, researchers with McKinsey & Company stated that "High uncertainty and low growth rates have forced companies in transportation, energy, manufacturing and other industries to squeeze every asset for maximum value."

The problem is that reactive maintenance exposes companies to significant risks and is not the transformational solution businesses need to remain or become competitive. Cheaper computational power, data streaming, autonomous data management and advanced analytics with embedded machine learning and visualization are enabling more efficient and effective asset utilization."

What is needed is a predictive maintenance system that relies on an informed approach for each production asset. It should gather data from multiple connected sources such as temperature and acceleration so that predicting failure is more regulated.

A key functional component of assets and equipment in many industries, and the topic of extensive analysis and Industry 4.0 (I4) focus, is the mechanical bearing. Bearings are critical components of rotating equipment including engines, fans, pumps and machines of all types. They are responsible for the continuous operation of planes, vehicles, production machinery, wind turbines, air conditioning systems and elevator hoists. The purpose of bearings is to reduce friction between moving parts and to enable continuous operation. The name comes from the notion of 'bearing' the load of a rotating shaft or sliding surface.

Bearings are designed for continuous and long use but each one has a finite lifespan and eventually will fail. Bearing failure causes the equipment it supports to cease operation, resulting in impact that ranges from inconvenient (a household fan stops running) to disruptive (a production line goes down) to catastrophic (a vehicle engine fails). Maintenance managers want to minimize risk and avoid unexpected service disruptions, particularly when the economic or human costs are high, but replacing bearings requires equipment downtime and has significant cost. They are constantly trying to achieve a balance between equipment up time and maintenance cost.

Three approaches to bearing maintenance are: (1) run to failure and replace, (2) perform maintenance at scheduled intervals based on observed aggregate historical norms, and (3) use condition monitoring. Bearing condition monitoring is based on wireless sensors embedded in bearings or located in host assemblies. It involves analyzing huge volumes of vibration data, isolating frequencies associated with the bearing geometry, calculating the spectrum view of the data, analyzing the spectrum and then comparing the spectrum to historical data.

Before they fail, bearings emit telltale signs of weakness resulting from excessive wear. These signs include increased vibration and higher operating temperature. The trick is to use data streams to anticipate time to failure and to lower the risk of downtime, while maximizing useful life. Handling that stream, storing all the historical data, and running the machine learning models is all part of the big data story.

"If you wait too long, you can destroy the shaft and the bearing. But do it too early and you lose money by replacing a bearing that can run longer," says author Alan S. Brown.

Bearing Failure

Advances in technology have made it possible to establish normal operating conditions by continuously monitoring the performance of each individual bearing, including vibration, temperature, torque and rotational speed, and to then use machine learning to process vast amounts of data. The result is the capability to find hidden patterns that represent potential failure scenarios. and to predict the remaining life of the bearing.

The combination of IoT, machine learning and analytics provides a solution for maintenance managers, enabling them to optimize machine life, manage costs and reduce the risk of damaging failure. "Machine learning… comes without the prejudices of engineers who look for problems only when they expect to see them," Brown notes.

Combining this capability with powerful, visual analytics provides real-time insight into bearing condition and empowers engineers to reduce cost, raise up time and lower risk.

According to Krishna Raman of Frost & Sullivan, "The adoption of the Industrial Internet of Things (IIoT)-based smart bearings, which can self-diagnose impending faults and failures, is expected to significantly increase in aerospace and defense, wind turbines, railway and automotive" segments. Bearing manufacturers are now looking at ways to leverage data and analytics to provide predictability rather than just metal components, and to "…catapult one of the world's oldest mechanical devices into the digital future."

Visit our website to learn more about how to apply Oracle Big Data to your IoT strategy.

Guest author, Jake Krakauer (@JakeKrakauer) is the Head of Product Marketing, Oracle Analytics

Join the discussion

Comments ( 2 )
  • updates on big data Thursday, September 20, 2018
    Thanks for sharing such a great information on Big Data Hadoop. We are expecting more article from you in future. We are also providing some information about Big Data on our website.
    updates on big data
  • sap fico training in vijayawada Friday, October 5, 2018
    Good and effective post. Cool information. Kindly know sap hana training in vijayawada
Please enter your name.Please provide a valid email address.Please enter a comment.CAPTCHA challenge response provided was incorrect. Please try again.