Skip to main content

Operations & Maintenance

The Role of AI in Asset Performance Management - Start with the Data You Have

In this series on artificial intelligence, I will help you navigate the AI methods out there that are applicable to asset performance management (APM). Additionally, this content will provide the insights you need to build a filter for the flood of solution offerings on the market that claim some sort of AI capability. In this installment, I will discuss discriminative AI, which is an approach that requires you to provide the input data, training and feedback to enable the tool to make predictions. 

 

Access to Data Isn’t the Problem 

The late 2000s to early 2010s was a period when organizations began recognizing the massive potential of collecting and analyzing large volumes of digital information. Beginning in the late 2010s, the focus shifted to connecting physical devices and machinery to the internet, enabling real-time data collection and smart monitoring. In our current era, the rapid advancement of AI technologies, particularly generative AI and large language models, has brought AI to the forefront of technological discourse. Having access to data isn’t a problem for most organizations. Data is being collected today in volumes we couldn’t have imagined a generation ago. There are many solutions on the market that enable you to apply AI/ML to make use of that data.

 

You Might Have Better Data Than You Think 

In his book "How to Measure Anything", Doug Hubbard, the inventor of the Applied Information Economics (AIE) method and the founder of Hubbard Decision Research (HDR), argues that people often underestimate the amount of useful information already available to them. He suggests that most decision-makers believe they need more data than they actually do to make meaningful decisions.  He also argues that people often ignore or discount existing information because it doesn't appear perfectly precise, when in reality, imperfect data can still provide valuable insights.

Artificial intelligence has shown us how we can take what we believe to be imperfect data and transform it into meaningful information we can make decisions withWith that said, how can we apply AI-enabled methods to take the information we have to produce the information we don’t have? One of the more well-established approaches to artificial intelligence is known as discriminative AI. This form of AI produces classifications, predictions, or decisions about input data. Unlike generative AI, like ChatGPT, you supply the input data, train it, simulate it and decide how confident you are in its predictions.  Discriminative “models are extensively used in applications like medical image analysis, autonomous driving, and industrial inspection”. 

 

AI Enables Pattern Protections Against Emerging Threats 

As we apply APM to help us improve the performance of our asset management systems, AI increases an organization’s ability to contain risks and vulnerabilities. For example, industry has long been collecting data from sensors to monitor assets for vibration and temperature. These are very specific signals about specific failure modesWith AI, we can train a model to look for patterns in time series data for failure modes we haven’t fully identified, giving us a more effective way to contain emerging threats and protect our assets. 

 

HxGN APM Visual AIUsing Data We Have 

As we consider the information we collect from thousands of data points, the limits that existed on how we can make use of it, before the emergence of AI, simply aren’t there anymoreFor instance, earlier systems struggled with nuanced pattern recognition. Modern discriminative AI can detect incredibly subtle patterns in data. AI models can automatically learn relevant features, often finding patterns humans might miss. Previous approaches often required significant processing times, but current discriminative AI can make predictions in milliseconds. 

HxGN APM is purpose-built with an intuitive discriminative AI solution for anomaly detection. It comes with a powerful machine learning engine which directly enables asset managers to minimize failure risks at scale.  HxGN APM enables users to visually build, test and deploy AI models in minutes. This training helps detect patterns and conditions which identify emerging risks and asset performance issues. This capability enables users to begin embedding their knowledge and decision-making against the predictions generated by the AI-enabled model.  

Curious to know more? See how HxGN APM empowers equipment and operations experts to capture and scale their knowledge. 

About the Author

Asset management domain expert committed to taking the fun and excitement out of asset management. Three decades of international standards, enterprise advisory, digital solutions, and implementation experience. Helped deliver asset management solutions to water services sector, electric utilities, power generation, process manufacturing, mining, chemicals, and fleet organizations on six continents. Marc is a contributing member to ISO Technical Committee 251 since 2010, representing the interests of the USA. He served leadership roles including of Chair of ANSI Technical Advisory Group, and first Convener of the International Standard ISO 55011, Guidance for development and application of public policy to enable asset management.

Profile Photo of Marc Laplante