Glossary

This list is here so the words Griffin uses are understandable to people at every level of an organization. It was first published on March 21, 2023.

Adivarent Control

The Purdue Model of ISA-95 has been the accepted picture of process control and enterprise integration since the mid-1990s. It does not easily account for edge computing devices and resources that now sit between the control system and the operator. Adivarent Control® is Griffin’s name for that layer. It assists the control system with complex tasks, and it assists operators with knowledge-based responses that need constant attention. ISA-95 at ISA.org. More on Adivarent Control.

Artificial intelligence (AI)

In this setting, AI covers the ways computers direct devices, such as actuators, to do work or make adjustments that a person would otherwise do. That requires seeing patterns and correlations, including ones that have not appeared before, and responding as the process or the data changes, with enough prediction to choose a sensible move. AI is used as advice, and it is used as real-time control.

Data scientists

A newer kind of technical expert, whose work is to explore complex or very large data with a range of analytical tools. That work used to belong to statisticians, engineers, and mathematicians. Many data scientists come from those fields. The size and complexity of the data have made new tools necessary if the goal is a clearer view of correlations, trends, diagnostics, and where process improvement should focus.

Data analytics

A set of mathematical techniques and algorithms for examining data in order to understand the process behind it. The tool depends on the goal. Broadly, the work falls into trends, diagnostics, predictions, prescriptive methods, and cognitive methods, including AI. Because it is a specialty, data scientists often do this analysis, though some of it, especially manufacturing data analytics, has been automated. Manufacturing data analytics. Smarter maintenance brochure (PDF).

Digital factory

A general term, used for a wide range of implementations. In Griffin’s use it means a centralized organization of digital data, used to build dashboards so people can follow production, analytics tools so they can investigate the process, and digital models for continuous evaluation and optimization of the process. Digital factory papers.

Edge computing

The term spread with the Industrial Internet of Things. In cloud computing, the work happens in the cloud, which brings latency and bandwidth limits. Edge computing puts computing resources closer to the data so analysis can happen in real time. The edge computer can still use the cloud for other services. In IIoT, the edge device does not only collect data. It processes the data. That makes the data more immediately useful to dashboards, data scientists, and people building real-time digital factory systems.

On-premise computing

Computing resources installed at the customer’s site. An on-premise system is chosen when the data is too sensitive to place with a third-party cloud provider, or when the internet itself is an unacceptable cyber risk.

IIoT

The Industrial Internet of Things is the use of smart sensors and actuators, networked together, in an industrial application. Control systems, working with edge computing devices, can analyze and process that data more thoroughly and improve how a manufacturing facility runs.

Manufacturing data analytics

Data analytics applied to operations in process and product manufacturing. It is used to protect quality, raise throughput, improve efficiency, and optimize production. The work may be done offline by data scientists, or it may be built into algorithms that change the production process for the conditions at hand. Smarter maintenance brochure (PDF).

Machine learning (ML)

A part of artificial intelligence whose aim is to let machines learn from data and, in many cases, imitate intelligent human behavior. AI is useful for complex tasks people perform. What makes machine learning attractive is learning those tasks with little or no first-principles model embedded in advance, and seeing relationships and trends across many variables quickly. That is where high-return operating improvements often start.

Particle swarm optimizer (PSO)

A search algorithm inspired by biology, used to explore many solutions inside a defined space. It needs an objective function that states the goal, or what a “most fit” solution is. The optimizer creates candidate solutions, called particles, and evaluates them repeatedly. How close each particle is to the objective is shared, which guides the swarm toward the more successful results and, ideally, toward an optimal solution. Random particles are included so the search is less likely to settle on a local optimum. See the particle-swarm tutorial.