---
title: "AI Uses Thermal Imaging to Predict Molded Part Quality"
id: "12035"
type: "post"
slug: "ai-uses-thermal-imaging-to-predict-molded-part-quality"
published_at: "2026-09-08T13:11:44+00:00"
modified_at: "2026-09-01T16:12:58+00:00"
url: "https://www.plasticsengineering.org/2026/09/ai-uses-thermal-imaging-to-predict-molded-part-quality-012035/"
markdown_url: "https://www.plasticsengineering.org/2026/09/ai-uses-thermal-imaging-to-predict-molded-part-quality-012035.md"
excerpt: "Infrared thermography and neural networks help predict injection-molded part quality by analyzing thermal patterns linked to cooling and warpage."
taxonomy_category:
  - "Artificial Intelligence"
  - "Automation"
  - "Editor's Choice Technical Paper"
  - "Education & Training"
  - "Equipment"
  - "Industry"
  - "Injection Molding"
  - "Materials"
  - "Mold & Die Making"
  - "Process"
  - "Resins"
  - "Sensors"
  - "Testing & Analysis"
  - "Trending"
taxonomy_post_tag:
  - "convolutional neural network"
  - "infrared thermography"
  - "Injection Molding"
  - "Injection Molding Defects"
  - "molded part quality"
  - "Neural networks"
  - "online quality monitoring"
  - "part mass prediction"
  - "predictive quality control"
  - "process monitoring"
  - "tensile strength prediction"
  - "thermal imaging"
  - "warpage prediction"
---

[Home](https://www.plasticsengineering.org/)
 » [News](https://www.plasticsengineering.org/news/)
 » [Trending](https://www.plasticsengineering.org/c/trending/)
 » [Artificial Intelligence](https://www.plasticsengineering.org/c/trending/artificial-intelligence/)
 » AI Uses Thermal Imaging to Predict Molded Part Quality

# AI Uses Thermal Imaging to Predict Molded Part Quality

Infrared thermal imaging captures surface temperature patterns and can help identify heat distribution, cooling behavior, and process variations in industrial environments. Courtesy of Faulhaber.

### Infrared thermography and neural networks help predict injection-molded part quality by analyzing thermal patterns linked to cooling and warpage.

Injection molding quality depends on the interaction among material response, processing conditions, mold design, and cooling behavior. Filling, packing, cooling, and solidification inside the mold directly influence key quality indicators, including part mass, tensile strength, and warpage.

**You can also read:** [Real-Time Melt Monitoring in Extrusion and Injection Molding](https://www.plasticsengineering.org/2026/04/real-time-melt-monitoring-in-extrusion-and-injection-molding-011167/)

[Recent research](https://doi.org/10.1016/j.jmapro.2024.07.021)
 explored infrared thermography as a tool for online quality prediction in injection molding. Researchers captured thermal images of molded parts and trained a convolutional neural network to associate temperature-field patterns with final product quality.

## **Why Thermal Fields Matter**

Many quality checks still happen after molding, through dimensional measurement, mechanical testing, or visual inspection. These methods can confirm whether a part meets specifications, but they often detect problems only after production has already generated defective parts. By then, the process may have continued under unstable conditions for several cycles.

Thermal imaging offers another way to monitor quality by capturing the surface temperature of molded parts shortly after ejection. The temperature field can show how evenly the part cooled and where heat remained longer. Hotter regions or uneven cooling patterns may point to differences in packing, shrinkage, crystallization, or residual stress development.

Thermal imaging adds value because it captures a process signal that changes with part quality. A stable temperature pattern can indicate consistent cooling, while unusual hot spots or uneven cooling may signal problems before standard inspection catches them. This makes thermography useful as an early warning tool during production.

## **How the Model Worked**

Temperature-field images from molded parts can capture spatial cooling patterns linked to mass, tensile strength, and warpage prediction. Courtesy of [InfraTec.](https://www.infratec-infrared.com/thermography/industries-applications/plastics-industry/#injection-molding)

The study relied on an online infrared camera system to collect temperature-field images from injection-molded parts. Product mass, tensile strength, and warpage deformation functioned as quality indicators. The experimental design varied processing conditions and linked each thermal image to the corresponding measured quality data.

Image preprocessing included segmentation and data augmentation. These steps made the thermal images more consistent and increased the training dataset. The convolutional neural network then analyzed the temperature data and identified links between thermal patterns and the selected quality indicators.

After hyperparameter optimization and five-fold cross-validation, the model achieved average relative errors of 3.48% for product mass, 3.60% for tensile strength, and 8.00% for warpage deformation.

## **What the Results Show**

The results indicate that thermal images can capture quality-relevant information that conventional scalar process signals may miss. Instead of relying only on machine settings or isolated sensor values, the model analyzed the spatial distribution of temperature across the molded part.

This spatial information matters because injection-molded parts rarely cool uniformly. Geometry, wall thickness, gate location, packing behavior, and mold temperature can create local thermal differences. These differences can influence shrinkage, mechanical performance, and warpage.

The study also found that temperature features near the product boundary strongly affected predicted quality. This suggests that edge regions and cooling fronts may carry important information about dimensional stability and mechanical response.

## **Implications for Processors**

In production, this approach could help detect quality deviations earlier in the molding cycle. If the model identifies a thermal pattern associated with excessive warpage or lower tensile strength, processors could adjust process conditions before scrap accumulates.

The method also supports a more direct link between process monitoring and final part performance. Infrared thermography gives the model a full temperature map of the part, not just one process value. CNN can then detect local patterns that may relate to warpage, strength, or part mass. This combination could help processors move from post-process inspection towards online quality monitoring.

## **Remaining Barriers**

This approach still requires careful implementation. Camera position, surface emissivity, part geometry, timing after ejection, and environmental conditions can affect image quality. Model accuracy also depends on the quality of the dataset and the range of processing conditions included during training.

A model trained on one material, mold, or part geometry may not perform reliably in another production setup. For industrial use, processors need proper calibration, consistent image acquisition, and validation under realistic production conditions.

## **Towards Predictive Quality Control**

Infrared thermography does not replace conventional quality testing, but it can strengthen process monitoring by detecting quality-relevant thermal patterns earlier. The research shows that temperature-field data can support quantitative prediction of molded-part quality when combined with deep learning.

For injection molding operations that face tight tolerances, high scrap costs, or complex part geometries, thermal imaging could become more than a diagnostic tool. It could become part of a predictive quality-control system that links cooling behavior, process stability, and final product performance.

By **[Maria Vargas](https://www.plasticsengineering.org/author/mariavargas/)** | September 8, 2026

##### [Maria Vargas](https://www.plasticsengineering.org/author/mariavargas/)

[+ postsBio ⮌](#)

María José Vargas is a mechanical engineer and MSc candidate in Materials Engineering and Nanotechnology at Politecnico di Milano. Her work focuses on environmental stress cracking in polyethylene, polymer failure behavior, plastics processing, and sustainable polymer applications.

### Other posts you could read

[August 25, 2026Carbon Black Pigments Create New Recycling Compliance Challenges](https://www.plasticsengineering.org/2026/08/carbon-black-pigments-create-new-recycling-compliance-challenges-011883/)

[September 4, 2026Thermoplastic Polyurethane Foams for Wearable Electronics](https://www.plasticsengineering.org/2026/09/thermoplastic-polyurethane-foams-for-wearable-electronics-012011/)

[August 26, 2026Transporting Plastics in a Low-Carbon Economy](https://www.plasticsengineering.org/2026/08/transporting-plastics-in-a-low-carbon-economy-011894/)

[August 12, 2026UK PFAS Plan Sets National Monitoring Framework](https://www.plasticsengineering.org/2026/08/uk-pfas-plan-sets-national-monitoring-framework-011865/)

[September 2, 2026Petrochemical Feedstock Strategy Shifts Beyond Cost](https://www.plasticsengineering.org/2026/09/petrochemical-feedstock-strategy-shifts-beyond-cost-011929/)

[August 31, 2026How Plastics Supply Chains Became Global Networks](https://www.plasticsengineering.org/2026/08/how-plastics-supply-chains-became-global-networks-011920/)

[September 1, 2026Just-in-Time Plastics Processing Faces New Supply Chain Limits](https://www.plasticsengineering.org/2026/09/just-in-time-plastics-processing-faces-new-supply-chain-limits-011924/)

[August 11, 2026PA6F Nanofiber Membranes Offer Sustainable PFAS Removal](https://www.plasticsengineering.org/2026/08/pa6f-nanofiber-membranes-offer-sustainable-pfas-removal-011858/)

## Share Your Thoughts [Cancel reply](/2026/09/ai-uses-thermal-imaging-to-predict-molded-part-quality-012035/#respond)

- [https://twitter.com/share?url=https://www.plasticsengineering.org/?p=12035&text=AI+Uses+Thermal+Imaging+to+Predict+Molded+Part+Quality&via=Plastics_Mag](https://twitter.com/share?url=https://www.plasticsengineering.org/?p=12035&text=AI+Uses+Thermal+Imaging+to+Predict+Molded+Part+Quality&via=Plastics_Mag)
- [https://www.facebook.com/sharer.php?u=https://www.plasticsengineering.org/?p=12035&t=AI+Uses+Thermal+Imaging+to+Predict+Molded+Part+Quality](https://www.facebook.com/sharer.php?u=https://www.plasticsengineering.org/?p=12035&t=AI+Uses+Thermal+Imaging+to+Predict+Molded+Part+Quality)
- [https://www.linkedin.com/shareArticle?mini=true&url=https://www.plasticsengineering.org/?p=12035&title=AI%20Uses%20Thermal%20Imaging%20to%20Predict%20Molded%20Part%20Quality&summary=Infrared+thermography+and+neural+networks+help+predict+injection-molded+part+quality+by+analyzing+thermal+patterns+linked+to+cooling+and+warpage.&source=Plastics+Engineering](https://www.linkedin.com/shareArticle?mini=true&url=https://www.plasticsengineering.org/?p=12035&title=AI%20Uses%20Thermal%20Imaging%20to%20Predict%20Molded%20Part%20Quality&summary=Infrared+thermography+and+neural+networks+help+predict+injection-molded+part+quality+by+analyzing+thermal+patterns+linked+to+cooling+and+warpage.&source=Plastics+Engineering)
- [https://www.reddit.com/submit?url=https://www.plasticsengineering.org/?p=12035&title=AI+Uses+Thermal+Imaging+to+Predict+Molded+Part+Quality](https://www.reddit.com/submit?url=https://www.plasticsengineering.org/?p=12035&title=AI+Uses+Thermal+Imaging+to+Predict+Molded+Part+Quality)
- [/cdn-cgi/l/email-protection#c7f8b4b2a5ada2a4b3fa97aba6b4b3aea4b4e782a9a0aea9a2a2b5aea9a0e7eae7868ee792b4a2b4e793afa2b5aaa6abe78eaaa6a0aea9a0e7b3a8e797b5a2a3aea4b3e78aa8aba3a2a3e797a6b5b3e796b2a6abaeb3bee1a5a8a3befaafb3b3b7b4fde8e8b0b0b0e9b7aba6b4b3aea4b4a2a9a0aea9a2a2b5aea9a0e9a8b5a0e8f8b7faf6f5f7f4f2](/cdn-cgi/l/email-protection#c7f8b4b2a5ada2a4b3fa97aba6b4b3aea4b4e782a9a0aea9a2a2b5aea9a0e7eae7868ee792b4a2b4e793afa2b5aaa6abe78eaaa6a0aea9a0e7b3a8e797b5a2a3aea4b3e78aa8aba3a2a3e797a6b5b3e796b2a6abaeb3bee1a5a8a3befaafb3b3b7b4fde8e8b0b0b0e9b7aba6b4b3aea4b4a2a9a0aea9a2a2b5aea9a0e9a8b5a0e8f8b7faf6f5f7f4f2)
- [#comments](#comments)