What Is AI for Industry?

From an engineer’s copilot to fully autonomous systems, artificial intelligence brings new opportunities to the factory floor.

What is AI for Industry?

Artificial intelligence (AI) has created an explosion of possibilities during the past year in every business sector worldwide. Primarily driven by an interest in generative AI and large language models (LLMs), AI’s breakthrough into the mainstream has also brought with it a great deal of hype. AI is not a singular, magical solution that will solve every problem. Instead, AI for Industry is a powerful group of technologies that brings new opportunities to the factory floor.

There are several solution areas for industrializing AI into operations. They include:

  • Energy Efficiency: Diagnosing and making suggestions for improving the performance of energy-intensive processes reduces power consumption;
  • Predictive Maintenance: Finding the right time to schedule regular maintenance prevents unexpected and costly shutdowns;
  • Engineering Copilot: Leveraging a natural language chat-based assistant based on a generative AI model provides use cases, writes computer code, or even aids in the development of AI models;
  • Anomaly Detection: Identifying rare items, events, or observations that deviate significantly from the rest of the dataset helps find even the toughest anomalies in process behavior; and
  • Predictive Quality: Determining the outcome of a batch based its parameters ensures use of the ideal batch profile and reduces production waste.

The release of OpenAI’s ChatGPT earlier this year has accelerated AI’s adoption. Understanding its capabilities and limitations for industry ensures that embracing AI solutions includes a realistic assessment of how they will address business needs.

Breaking Barriers for Industry

The Main Categories of AI Solutions

At its core, AI refers to machines or systems that mimic human intelligence to perform tasks. They can improve themselves based on the information they collect.

There are generally four categories that represent AI for industry.

  1. Assisted Intelligence is the most basic form of AI. It helps engineers by making their tasks easier. For instance, assistive intelligence might use operational data to suggest an improvement. However, the final choice is up to the engineer. Assistive intelligence devices cannot make decisions on their own.
  2. Automation intelligence solutions, on the other hand, are suitable for repetitive tasks that do not require much human interaction. An example is a machine that sorts products on a conveyor belt.
  3. Augmented Intelligence works with humans to perform tasks better than either could alone. An augmented intelligence solution could analyze sensor-generated data to help engineers make better and quicker decisions.
  4. Autonomous intelligence is the most advanced form of AI. It can make decisions and act on its own without human help. A self-driving vehicle in a warehouse uses autonomous AI.

Autonomous solutions, however, are not yet suitable for managing manufacturing processes because they can act on their own without human assistance. The level of trust in autonomous systems is too low to allow them to make changes on the factory floor. The three remaining categories are a set of services, functions, models, and techniques that, when put together, emulate human intelligence. The most useful of these categories for operations are deep learning, machine learning (ML), and LLMs.

AI generally can be broken down into four categories. Assisted, automation, and augmented intelligence show the most promise for the process manufacturing industry right now.

AI for Industry: Machine Learning

Machine learning, which is a subset of AI, uses algorithms to learn from and make selections based on underlying data. Most data scientists working in the process manufacturing industry use ML, and its subset, deep learning, more than other techniques. Python is the programming language of choice for sorting data and developing machine learning models. Popular Python libraries, such as SciKit Learn, PyTorch, and Tensorflow, offer access to these techniques with a simple installation package.

Machine learning applies mathematical algorithms to data. These algorithms then generate estimates using unsupervised or supervised models. Supervised models require a labeled dataset, while unsupervised models discover general patterns in data. ML models can make estimations based on historical data, categorize events, and cluster data points based on commonalities.

With machine learning, computers can learn from data without being explicitly programmed to do so. They are used for descriptive, predictive, or prescriptive functions.

Anomaly detection model

Applications of ML for Industry

The choice of which model to use for a given exercise is based on the experience of the data scientist. Examples of their application include:

  • Estimations of a pump’s pressure, flow, and vibration sensor, with an unsupervised anomaly detection model trained on time-series data. Clustering or dimension reduction models are often used here.
Decision tree model

A decision tree model, such as the one shown above, can be used to understand the performance of a process.

  • Understanding the performance of a process. Supervised models, such as a linear regression or decision tree, tell the algorithms how they should perceive the data. A pump’s performance status might be good, bad, or offline for maintenance. The latter requires the use of contextual data from various operational systems in addition to sensor-generated data.
  • Learning by reinforcement, such as Q-learning. In these situations, data scientists train an algorithm with rewards or penalties for taking or suggesting a certain action. Self-driving cars use reinforcement learning, but it is useful for optimizing industrial processes when combined with the expertise of operational experts.
  • The development of soft sensors. Soft sensors could serve as additions to digital counterparts or take their place in harsh environments, for example. These require a supervised model, such as a neural network or gradient boosting algorithm.
  • These can predict or classify batch quality or identify equipment failure. Classification models often use Naïve Bayes or decision tree algorithms.
Soft sensor design

Into the Future with Generative AI

AI techniques, including machine learning, have been used in the industry for many years. But generative AI is experiencing its peak of rapid advancement. A breakthrough in transformers has paved the way for innovations and new releases, including ChatGPT. In just over a year since the announcement of the breakthrough, several major tech companies have launched their own LLMs. These include Google’s Palm 2 and Meta’s Llama 2.

2023 Gartner Hype Cycle

The 2023 Gartner Hype Cycle for Emerging Technologies shows Generative AI is at the peak of its inflated expectations. Courtesy of Gartner.

While Generative AI is at the peak of the 2023 Gartner Hype Cycle for Emerging Technologies, Operational experts already can use the new technology for a variety of purposes. In the next blog, we will explore the explosion of Generative AI and its use in the factory.

AI ebook CTA

From our Blog

The Organizational Gap

Bridging the Organizational Gap in Pursuit of an Augmented Factory

By closing the Organizational Gap, manufacturers can do more to improve operations with machine learning models and AI-powered systems.

Achieve Better Operational Storytelling with a Connected Factory

A Connected Factory unlocks data silos and provides a more holistic view of operations to help engineers make better decisions.
The smart factory with the help of IT OT Convergence

Navigating the IT/OT Convergence in Pursuit of a Smart Factory

The IT/OT convergence facilitates a more innovative, efficient, and adaptable manufacturing environment in a digital transformation.