A recent research combines Artificial Neural Networks and Genetic Algorithms in a closed-loop predictive technique to boost product quality by accurately forecasting performance and optimizing assembly decisions in real-time manufacturing.
In today’s fast-paced manufacturing world, every second ⏱️ and every component ⚙️ counts. The cost of producing defective products or wasting resources can be huge. That’s where predictive techniques come in—they help manufacturers foresee potential performance issues early in the design or assembly stage, before defects happen.
Traditionally, industries relied on statistical quality control—checking samples after production and fixing problems reactively. But modern factories need smarter, faster, and proactive methods that catch quality issues at the root.
The research we’re exploring today, presents a closed-loop system that marries two powerful tools from artificial intelligence:
Together, they create a predictive–optimization duo that not only forecasts product performance but also actively improves assembly quality in real-world manufacturing.
Imagine teaching a system to recognize the hidden rules that decide whether a product will perform well. That’s what ANNs do.
In this research, the ANN was trained to predict the performance of hermetic compressors (used in refrigeration systems). These machines are very sensitive—tiny differences in part dimensions can make or break efficiency.
These four factors were fed into the ANN. By running 32 carefully designed experiments (using the Design of Experiments method), the researchers created a dataset.
The ANN model architecture looked like this:
🎯 The results?
In short, the ANN could reliably predict compressor performance just from input dimensions, without the need for lengthy testing.
Prediction alone isn’t enough. You also need to decide: Which combination of parts will give the best final product? That’s where Genetic Algorithms (GAs) come in.
GA works like evolution in nature:
In this study, the GA used the ANN’s predictions as fitness criteria. Essentially, ANN acted as the evaluator, while GA searched through possible part combinations until it found the best assembly configurations.
The GA was programmed to maximize the number of Class A parts in final assemblies. It ran thousands of “generations” until the best results emerged.
The closed-loop ANN–GA system produced remarkable improvements:
Not only that, the whole process was fast:
This makes it practical for real-world smart factories where quick decisions are essential.
A big innovation in this study is the closed-loop design:
This creates a self-improving cycle—a system that learns and adapts continuously.
Compared to older open-loop methods (where ANN and GA work separately), this integration makes decisions more accurate, adaptable, and realistic for industrial settings.
The study shows a strong proof-of-concept, but the possibilities go way beyond compressors:
In essence, ANN–GA frameworks could become the standard backbone of Industry 4.0 quality control systems.
This research is an exciting leap in how we think about predictive quality management. By combining brain-inspired prediction 🧠 with evolution-inspired optimization 🧬, manufacturing can move from reactive problem-solving to proactive excellence.
The dream? A future where every product rolling off the line is optimized, efficient, and high-quality—by design 🌟.
Predictive Technique - A smart method that uses past data 📊 and models to forecast future outcomes—like predicting product performance before it’s even built.
Product Quality - How well a product performs compared to what it was designed to do 💡—covering durability, efficiency, and customer satisfaction.
Artificial Neural Network (ANN) - A computer model inspired by the human brain 🧠 that learns patterns from data to make predictions, like guessing performance from part dimensions. - More about this concept in the article "Smarter, Stable Smart Grids ⚡ Hybrid AI".
Genetic Algorithm (GA) - An optimization method inspired by evolution 🧬—it tries many “solutions,” keeps the best ones, and mixes them until the best design is found. - More about this concept in the article "Digital Twin Boosts Vertical Farming 🌱".
Closed-Loop System - A feedback-driven setup 🔄 where predictions and optimizations continuously improve each other instead of working separately. - More about this concept in the article "AUV Solar Optimization 🌊 The Next Wave in Marine Robotics".
Design of Experiments (DOE) - A structured way to test different factors systematically 🧪—helping scientists figure out which variables really matter.
Performance Index (PI) - A numerical score 📏 that measures how well a product performs compared to its design target.
Optimization - The process of finding the best possible solution among many choices ✅—in this case, the best combination of parts for top product quality. - More about this concept in the article "Harnessing Nature: How Harris Hawks Optimization Is Revolutionizing Power Grids 🦅 ⚡".
Smart Manufacturing - The modern approach to production 🏭 where AI, sensors, and data-driven techniques make factories faster, smarter, and more efficient.
Source: El-Baz, M.A.; Abd-Elwahed, M.S. Enhancing Product Quality Using Artificial Neural Networks and Genetic Algorithms. J. Manuf. Mater. Process. 2025, 9, 322. https://doi.org/10.3390/jmmp9090322
From: King Abdulaziz University.
