INFORM April 2026

14 INFORM APRIL 2026 , VOL. 37, NO. 4

AI TOOLS FOR PROCESS CONTROL AI tools for process optimization and automation span a wide range of methods and capabilities. Among them, machine learning (ML) approaches are especially effective at analyzing large experimental datasets, identifying patterns, and making predictions or classifications. Artificial neural networks (ANN), one of the earliest types of ML, remain a workhorse for process optimization. Inspired by the structure of the human brain, ANN consist of interconnected layers of computational units, or artificial neurons, that transmit signals to one another. Each neuron processes inputs and sends outputs to other connected neurons, while weighted connections determine how strongly one neuron influences another. During training of an ANN, connection weights are iteratively adjusted to

Input Layer

Intermediate Layers

Output Layer

Scheme of a multilayer perceptron, the most widely used type of artificial neural network in bioinformatics prediction models. Source: Gonzalez-Fernandez, I., et al., Crit. Rev. Food Sci. Nutr. 59 (2018)

minimize error between predicted and actual outcomes, which enables the AI to learn complex, nonlinear relationships. Some optimization algorithms, including particle swarm optimization (PSO) and genetic

algorithms, are also inspired by biology. For example, PSO mimics the collective behavior of animals—flocks of birds, schools of fish, or swarms of insects—as they search for food. PSO searches large, complex spaces of variables to identify optimal outcomes,

Examples of AI tools for process optimization and automation in the fats and oils industry. Sources: in table.

Sample References

Category

Examples

General Mechanism

Sample Applications

Machine Learning ANN, Random Forest, XGBoost, CatBoost, MTDL

Learn nonlinear relationships from historical or experimental data to make predictions or classifications Optimize process parameters directly or in combination with ML models Simulate factory behavior or use predictive models to adjust variables in real time Combine sensor data with ML to interpret real-time signals for classification and defect detection

Predict biofuel yield; detect adulterants; predict catalyst performance; conduct life cycle assessments Optimize reaction conditions; detect adulterants Virtual testing; adaptive reactor control; predictive maintenance

1 , 2 , 3 , 4

Optimization Algorithms

PSO, GA

1 , 5

Tools for Process Simulation and Real Time Control

Digital twins, ML-augmented MPC, RL

6

Sensor-Integrated AI IoT sensors + ML; CNN-based vision systems

Fruit grading; impurity detection; feedstock characterization

2