As we’ve talked about earlier, industrial companies now have the ability to interact with physical equipment like never before. You’ve probably also heard that there’s an urgency for you to get started with your digital transformation. However, you might be bombarded with different terms you don’t understand. Here I’ll give you definitions of some terms we often use in relation to digital transformation in heavy-asset industries.


Advanced analytics

Advanced analytics is the collection of sophisticated modeling techniques used to investigate data or content to gain deeper insights, predict future events, discover patterns and generate recommendations. Some of the different methods or techniques within advanced analytics include:

  • data/text mining
  • descriptive modeling
  • statistical/quantitative analysis
  • machine learning
  • pattern matching
  • forecasting
  • semantic analysis
  • sentiment analysis
  • visualization
  • network and cluster analysis
  • multivariate statistics graph analysis
  • simulation and optimization
  • neural networks


Anomaly detection

Anomaly detection (a.k.a. outlier detection) is a class of problems, where the goal is to determine behavior that’s sufficiently different from the nominal (or normal) behavior patterns. In the context of industrial settings, anomalies are unexpected or undesired behavior. Typically, anomalies occur rarely in well-behaved systems. Anomaly detection is one of the core techniques in data-driven maintenance.


Artificial intelligence (AI)

Artificial intelligence is the ability of a machine to perform tasks that would otherwise require human-like intelligence in some form. Common approaches to artificially intelligent systems are machine learning-based or rules-based systems, with machine learning being the currently favored approach.


Computer vision

Computer vision is how we can enable computers to "see" and gain understanding of digital images in the same way humans do (or even surpass humans). The goal is to extract meaning from pixels. Computer vision is acquiring, processing, analyzing and understanding these digital images and extract high-dimensional data from the real world to produce numerical or symbolic information, e.g. in the form of decisions.


Deep learning

Deep learning refers to a type of machine learning that uses artificial neural networks. Deep neural networks are "deep" because they are composed of many layers, allowing them to extract patterns in the data in a hierarchical way. Deep learning has excelled at tasks related to images and language, among others.


Edge analytics

Edge analytics is an approach where you collect and analyze data at a non-central point, such as the sensor, device or touchpoint itself. Analyzing data as it’s generated lowers companies’ costs and increases the ability to drive timely, valuable business insights from all of the data to inform actions and decisions at remote sites.

Want to learn more about edge analytics? Check out our whitepaper Edge Analytics and the Industrial IoT:

Download our Whitepaper Edge Analytics and the IIoT


Edge computing

Edge computing refers to "on device" computing, rather than cloud-based computing. In our daily lives, we see edge computing that happens on our smartphones. Analytics for the industrial internet of things (IIoT) is boosted by edge computing.


First-principles model

First-principles models are based on physical and chemical properties and well-understood engineering principles. They often underlie existing simulation and control of physical systems and can be combined with data-driven models, such as machine learning, to improve speed and accuracy of hybrid analytics.


Industrial Internet of Things (IIoT)

The industrial internet of things is the extension and use of the internet of things (IoT) in industrial applications. It’s the use of smart sensors and connectivity allowing data collection, exchange, and analysis that’s enabling optimized productivity and efficiency.


Machine learning (ML)

Machine learning is a subset of AI and is a class of predictive algorithms that learn representations of data sets they’re provided with. The algorithms build a mathematical model based on training data in order to make predictions or decisions without being explicitly programmed to do a certain task. There are different types of learning algorithms such as supervised learning, unsupervised learning, and reinforcement learning.

Machine learning can create value for industrial companies in three ways:

  1. improving their operational performance
  2. optimizing their systems and processes, and/or
  3. create new business models


Neural network

Artificial neural networks are a class of machine learning algorithms designed to make predictions. Like other machine learning algorithms, they can be used for several types of prediction problems, such as predicting the class of data (e.g. the type of rock in a sample) or a numerical value (e.g. the expected lifetime of an asset). Deep neural networks (a.k.a. deep learning) is an advanced type of neural network. Neural networks are good at extracting useful patterns from data in a way that other machine learning methods are less capable of.


Predictive equipment maintenance

True predictive equipment maintenance involves forecasting the specific mode of pending equipment failure. Predictive equipment maintenance for critical equipment is a key goal for many industrial companies. However, in order for a machine to recognize all potential failures, it must see a pattern of each possible failure many times for that specific equipment.

For equipment that rarely fails, rather than trying to predict failures from the outset, at Arundo, we start by identifying unusual patterns of behavior in the underlying data using anomaly detection. This approach is generally much more actionable and realistic for operators, as it allows diagnostics on specific sensors and operating conditions.

Want to learn more about the journey to predictive equipment maintenance? Read this blog article.