Composite Curing Temperature Profile Control Using Thermal Stack ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In composite part curing processes, existing methods face challenges in accurately controlling temperature profiles due to complexities in heat transfer boundary conditions, especially when multiple parts are cured together, leading to issues like under-cured areas, porosity, and residual stresses, as traditional numerical simulations and thermocouple measurements are unreliable.
Innovation Solution
A machine learning-based approach is employed to determine thermal stack parameters and optimize temperature profiles by using sensor data from tool and ambient temperatures within a heating vessel, allowing for real-time adjustment of heating operations to meet process specifications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional numerical simulation and thermocouple measurements are used to control temperature profiles, then the curing process can be monitored, but the measurements become unreliable when multiple parts are cured together due to complex heat transfer boundary conditions
Solution Approach 1:
The patent introduces machine learning models as intermediary systems that indirectly infer the complex heat transfer boundary conditions and part temperatures. Instead of directly measuring difficult-to-access temperatures with thermocouples, the ML models use readily available sensor data (air temperature, tool temperature) to predict part temperatures and boundary conditions, effectively mediating between simple measurements and complex thermal behavior
Solution Approach 2:
The patent replaces the physical thermocouple measurement system with a computational machine learning-based temperature prediction system. The ML models substitute the mechanical/physical measurement approach with an information-processing approach, using algorithms to predict temperatures based on patterns learned from training data, thereby avoiding the limitations of physical sensor placement
2Measurement precision
If thermocouples are placed at lagging and leading locations to monitor part temperature history, then temperature specifications can be verified, but access to these locations is impeded when parts are coupled to tools
Solution Approach 1:
The patent uses machine learning models as intermediaries to infer part temperatures at critical locations without requiring physical sensors at those locations. The models mediate between easily accessible measurements (tool temperature, air temperature) and the difficult-to-access part temperature data, eliminating the need for intrusive thermocouple placement
Solution Approach 2:
The patent creates virtual copies of the temperature measurement function through machine learning models. Instead of physically placing thermocouples at lagging and leading locations, the ML models generate virtual temperature readings that replicate what thermocouples would measure, based on patterns learned from training data and readily available sensor inputs
3Productivity
If multiple parts are cured together in an autoclave or oven, then productivity increases, but convective airflow patterns change in complicated ways that are difficult to model
Solution Approach 1:
The patent implements feedback loops where machine learning models continuously predict temperatures and boundary conditions based on current sensor readings, compare predictions with actual measurements, and use this feedback to improve subsequent predictions. This feedback mechanism allows the system to adapt to the complex, changing airflow patterns that occur when multiple parts are cured together, maintaining measurement accuracy despite increased productivity
Solution Approach 2:
The patent replaces complex physical modeling of convective airflow with machine learning-based prediction. Instead of attempting to model the complicated fluid dynamics of hot air flow around multiple parts, tools, and fixtures, the ML models learn the effective thermal behavior from training data, substituting mechanistic modeling with data-driven prediction that naturally captures the complex airflow patterns
4Manufacturing precision
If heat transfer boundary conditions are unknown due to tool nesting and orientation, part geometry, and overall thermal mass, then numerical simulation becomes inaccurate, but obtaining these parameters increases measurement complexity
Solution Approach 1:
The machine learning models perform self-service by automatically learning and inferring the heat transfer boundary conditions from training data without requiring explicit measurement or input of these parameters. The models internally capture the effects of tool nesting, orientation, part geometry, and thermal mass through patterns learned during training, eliminating the need for separate measurement campaigns to obtain these difficult-to-measure parameters
Solution Approach 2:
The patent substitutes direct measurement and numerical modeling of heat transfer boundary conditions with machine learning-based inference. Instead of physically measuring or calculating the complex boundary conditions arising from tool nesting and part geometry, the ML models predict their effects on part temperatures through patterns learned from training data, replacing complex physical measurement and modeling with computational prediction
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This method effectively identifies and adjusts temperature profiles to ensure composite parts meet process specifications, reducing defects and improving curing efficiency by leveraging machine learning to model complex heat transfer dynamics.
Implementation Method 1
numeric process simulation of the thermo-chemical curing reaction
Implementation Method 2
convective thermal resistance
Implementation Method 3
Composite parts are often heated via convection in ovens or autoclaves
Implementation Method 4
conductive thermal resistance
Implementation Method 5
after an exothermic curing reaction starts in the composite part, part temperature at the center of the part may be greater than the air temperature
Data Source
AI summary
Heating operation control includes obtaining sensor data indicating measured temperatures within a heating vessel during a heating operation; determining sets of thermal stack parameters. Each set of candidate thermal stack parameters is descriptive of a respective configuration of a thermal stack modeled by a first machine learning model to generate one or more estimated tool temperature values. The thermal stack includes the tool and a part coupled to the tool. Heating operation control also includes determining a temperature profile for the heating operation. The temperature profile is determined, via a second machine learning model, based on the plurality of sets of thermal stack parameters and one or more process specifications of the thermal stack.