Robustness-performance optimisation for automated composites manufacture
Lead Research Organisation:
Cranfield University
Department Name: School of Water, Energy and Environment
Abstract
This project focuses on the development of a manufacturing route for composite materials capable of producing complex components in a single process chain based on advancements in the knowledge, measurement and prediction of uncertainty in processing. The methodology proposed uses measurements of the instantaneous state of a component during production, predictive modelling of associated variability and numerical optimisation. These three are integrated in a control loop that allows the process to adapt in real time in order to compensate for deviations from its nominal state due to variability. This manufacturing philosophy accepts the existence of variability in these highly heterogeneous and directional materials and uses it in order to improve the product as the process evolves.
The necessary developments comprise major manufacturing challenges, such as the real time measurement of fibre variability in robotic fibre placement and the processing of composite components involving areas of large thickness. These are accompanied by significant mathematical advancements, such as the numerical solution of coupled non-linear stochastic partial differential equations, the inverse estimation of composite properties and their probability distributions in different directions based on real time measurements and the formulation and solution of a stochastic model of the variability in fibre arrangements. The integration of these developments will be carried out on a single process chain of fibre placement, resin infusion and resin cure; however their applicability is generic in the context of manufacturing involving heterogeneous materials and variability.
The outcome of this work will enable a step change in the capabilities of composite manufacturing technologies to be made, overcoming limitations related to part thickness, component robustness and manufacturability as part of a single process chain, whilst yielding significant developments in mathematics with generic application in the fields of stochastic modelling and inverse problems.
The necessary developments comprise major manufacturing challenges, such as the real time measurement of fibre variability in robotic fibre placement and the processing of composite components involving areas of large thickness. These are accompanied by significant mathematical advancements, such as the numerical solution of coupled non-linear stochastic partial differential equations, the inverse estimation of composite properties and their probability distributions in different directions based on real time measurements and the formulation and solution of a stochastic model of the variability in fibre arrangements. The integration of these developments will be carried out on a single process chain of fibre placement, resin infusion and resin cure; however their applicability is generic in the context of manufacturing involving heterogeneous materials and variability.
The outcome of this work will enable a step change in the capabilities of composite manufacturing technologies to be made, overcoming limitations related to part thickness, component robustness and manufacturability as part of a single process chain, whilst yielding significant developments in mathematics with generic application in the fields of stochastic modelling and inverse problems.
Planned Impact
The main outcome of this project will be the capability to produce highly performing composite structures in a single process chain. The load bearing capabilities of composites will be increased whilst their cost will be reduced. Consequently, use of composites as a lighter, greener, less costly solution will result in significant environmental benefits, business opportunities and increased competitiveness. More generic application of the project outcomes will also enable a shift in how manufacturing addresses variability in general, with associated benefits in terms of production speeds and final product quality. This will lead to increased efficiency and result in a competitive advantage for UK manufacturing companies.
The impact of the work will be maximised through direct industrial collaboration in the context of the project. Furthermore, the opportunities for long term benefits will be enhanced by a communications strategy addressed to industry and the academic community thorough workshops, conference presentations and publications in scientific journals.
The impact of the work will be maximised through direct industrial collaboration in the context of the project. Furthermore, the opportunities for long term benefits will be enhanced by a communications strategy addressed to industry and the academic community thorough workshops, conference presentations and publications in scientific journals.
Publications
Crowley D.M.
(2016)
Hand lay-up of complex geometries-prediction, capture and feedback
in International SAMPE Technical Conference
Elkington M.
(2016)
Automated Layup of sheet prepregs on complex moulds
in International SAMPE Technical Conference
Matveev M.Y.
(2016)
Dual flow front measurements for improved permeability characterisation
in ECCM 2016 - Proceeding of the 17th European Conference on Composite Materials
Warburton J
(2016)
Digital Image Correlation Vibrometry with Low Speed Equipment
in Experimental Mechanics
Mesogitis T
(2016)
Stochastic heat transfer simulation of the cure of advanced composites
in Journal of Composite Materials
Matveev M
(2016)
Understanding the buckling behaviour of steered tows in Automated Dry Fibre Placement (ADFP)
in Composites Part A: Applied Science and Manufacturing
K. Tifkitsis
(2017)
STOCHASTIC MULTI-OBJECTIVE OPTIMISATION OF COMPOSITES MANUFACTURING PROCESS
K. Tifkitsis
(2017)
On-line uncertainty estimation in composites manufacturing
Matveev M
(2017)
Uncertainty in geometry of fibre preforms manufactured with Automated Dry Fibre Placement and its effects on permeability
in Journal of Composite Materials
Park M
(2017)
Stochastic Resin Transfer Molding Process
in SIAM/ASA Journal on Uncertainty Quantification
| Description | Evaluation of variability in the constituents of composite materials and how these propagate through the manufacturing process: The influence of fibre variability on the behaviour of the material during the process as well as the generation of defects, such as residual stress has been established. The effect of uncertainty in the initial state of the polymer matrix on the generation of temperature overshoot has been evaluated; whilst the role of variability in the process conditions imposed during the manufacturing of composite components (temperature of tooling, cooling through convection) has been established. The associated methods, comprising both experimental tests and simulation capabilities, have been developed. In the case of high performance aerospace grade composites the residual stress can vary by more than 10% due to uncertainty in fibre orientation, whilst the uncertainty in tool temperature alongside the initial chemical state of the thermosetting resin and the activation energy of its cross-linking reaction are the dominate variations in the final product. Estimation/approximation of the trade-off between process quality and cost through multi-objective optimisation: The manufacturing process has been investigated in terms of the interplay between process cost and the probability of deviations from nominal quality through unpredictable temperature deviations. A methodology based on Genetic Algorithms has been developed to approximate the corresponding trade-off surface and the problem of challenging curing of thick and ultra-thick components ha been addressed. It was shown that the methodology can lead to the adoption of process design that allow process robustness at similar levels with conventional/current conditions at a fraction of the cost associated with the duration of the process. Benefits in the range of 30-70% have been achieved. Efficient stochastic simulation methods for complex, high dimensionality variability: The complexity of the problems associated with composites manufacturing has been addressed with the development of tailored methodologies able to deal with the coupled physics associated with the process of cure (advection-reaction-diffusion), such as stochastic collocation and Wiener chaos constructed with a recursive multistage algorithm and the large number of random variables met in the flow through porous media problem of mould filling addressed using multilevel Monte Carlo and sampling on irregular grids. In addition a highly efficient method based on surrogate modelling using Kriging has been developed and implemented. Its efficiency has been tested in the manufacturing of fibrous composites using Resin Transfer Moulding and work is being carried for the adoption of this method to allow targeted non destructive evaluation of composite structures based on the outcome of process monitoring and stochastic simulation. Integration of inverse problem solution with on-line monitoring of the manufacturing process. The integration of flow monitoring based on dielectrics with inverse solutions of the corresponding flow and cure problem has been accomplished and demonstrated off-line. The capability of estimating unknown parameters , such as permeability, with a lower level of uncertainty has been demonstrated. |
| Exploitation Route | Industrial benefits of the research outcomes of the work include: -Use of process optimisation for thermal conditions design with significant reduction of process duration and associated costs (30-70% in the case of thick/ultra-thick components). -Incorporation of explicit uncertainties in design resulting in less conservative solutions. Quantification of the potential benefit will be possible upon completion of the integration of optimisation/stochastic simulation tools. Research benefits are as follows: -Development of a set of methods able to cope with the complexity of spatial variability and coupled physics related to composites manufacturing. These address the combination of continuous fibre variability and scalar variations observed in matrix chemical state and thermal conditions with the free boundary flow problem and the reaction-diffusion associated with curing and can be expanded in a number of different research areas. -Development of efficient surrogate models that allow incorporation of non linear coupled manufacturing simulation within a stochastic optimisation procedure. -The quantification of actual input variability and its representation in stochastic objects provides the basis for variability analysis of current and future variations of process strategies used in composites manufacturing. |
| Sectors | Aerospace, Defence and Marine,Energy,Transport |
| Description | The project generated a sensing concept and design for composite manufacturing able to track the flow and cure of resin during the manufacturing of thermosetting matrix carbon fibre composites, documented in one of the project outputs (KI Tifkitsis, AA Skordos. A novel dielectric sensor for process monitoring of carbon fibre composites manufacture. Composites Part A: Applied Science and Manufacturing 2019;123:180-189). The attractive features of the sensing technology developed were its applicability with electrically conductive reinforcement such as carbon, its accuracy, its simplicity in installation and its flexibility in terms of geometry and moulding configurations. Due to their advantages, the sensors established in this work have been selected for further development by the National Composites Centre through a funded High Value Manufacturing Catapult Technology Pull Through project involving the PI. This matured the TRL of the sensors and demonstrated their industrial applicability (Contact: Jack Alcock, Advanced Technology Project Lead, National Composites Centre, Jack.Alcock@nccuk.com). In the context of this development, a specialist equipment company (ADVISE-DETA Ltd.) specialising in composites manufacturing process monitoring and control developed specialised connectors for in mould use of the sensors and adapted its data acquisition software for integration of the sensors developed (Contact: George Maistros, Technical Director, ADVISE-DETA Ltd., gmaistros@advise-deta.com). The application of the sensors was taken further by the NCC in collaboration with Meggitt Aerospace (Contact: Peter Karapappas, Principal Engineer, Airframe Systems, Meggitt Aerospace Company, Peter.Karapapas@meggitt.com), with testing and utilisation of the technology in an industrial context during Resin Transfer Mounding of a composite housing for a commercial aeroengine heat exchanger. The features highlighted by this activity demonstrated that the sensors developed in RPOACM provide an attractive solution for digitisation of composites manufacturing combining accuracy in measuring flow and cure on line with simple installation, non-invasiveness, low cost and flexibility in use and operation. The impact story around this work featured in a long article (https://www.compositesworld.com/articles/national-composites-center-and-meggitt-demonstrate-linear-dielectric-sensor-for-liquid-composites-processing), a short article (https://www.compositesworld.com/news/ncc-sensor-technology-unlocks-digital-potential-for-liquid-resin-processes) and a technology overview (https://www.compositesworld.com/articles/sensors-data-for-next-gen-composites-manufacturing) in Composites World and in a news story by the NCC (https://www.nccuk.com/news/sensor-technology-unlocking-digital-potential-in-composites-manufacturing/) achieving further dissemination to the wider composites community. |
| First Year Of Impact | 2021 |
| Sector | Aerospace, Defence and Marine |
| Impact Types | Societal,Economic |
| Description | Feasibility study |
| Amount | £50,000 (GBP) |
| Funding ID | Active control of the RTM process under uncertainty using fast algorithms |
| Organisation | Engineering and Physical Sciences Research Council (EPSRC) |
| Sector | Public |
| Country | United Kingdom |
| Start | 02/2018 |
| End | 07/2018 |
| Description | Flexible RTM tool with automated distortion correction - H2020, Clean Sky 2 |
| Amount | € 1,400,000 (EUR) |
| Funding ID | 821488 |
| Organisation | European Commission |
| Sector | Public |
| Country | European Union (EU) |
| Start | 09/2018 |
| End | 08/2020 |
| Description | Simulation tool development for a composite manufacturing process default prediction integrated into a quality control system - H2020, Clean Sky 2 |
| Amount | € 695,000 (EUR) |
| Funding ID | 686493 |
| Organisation | European Commission |
| Sector | Public |
| Country | European Union (EU) |
| Start | 01/2016 |
| End | 01/2019 |
| Title | A novel dielectric sensor for process monitoring of carbon fibre composites manufacture |
| Description | -Flow sensor validation: includes the lineal sensor results during RTM processing and comparison with visual monitoring data-Cure sensor: includes the cure sensors results of isothermal runs of neat resin and VART and the comparison with existing models. |
| Type Of Material | Database/Collection of data |
| Year Produced | 2019 |
| Provided To Others? | Yes |
| Impact | New sensor able to to monitor continuously filling progress in liquid moulding of composites and the associated validation. |
| URL | https://cord.cranfield.ac.uk/articles/A_novel_dielectric_sensor_for_process_monitoring_of_carbon_fib... |
| Title | Measurement of thermal conductivity of epoxy resins during cure |
| Description | - -Thermal_conductivity_glycerine: Experimental and FE data of the thermal conductivity test run on glycerine at 100 degC and 4 degC. Includes also the sensitivity analysis of the measurement on thermocouple misplacement. - -RTM6_cure_kinetics_results: Includes the isothermal and dynamic data for the RTM6 cure kinetics characterisation and the corresponding model. It also includes the sensitivity analysis on period and amplitude influence for RTM6 - - 890RTM_cure_kinetics_results: Includes the isothermal and dynamic data for the 890RTM cure kinetics characterisation and the corresponding model. It also includes the sensitivity analysis on period and amplitude influence for 890RTM - - XU3508_cure_kinetics_results: Includes the isothermal and dynamic data for the XU3508 cure kinetics characterisation and the corresponding model. - -Di Benedetto_XU3508: Includes the dataused to build the Di Benedetto equation for the XU3508 resin system and the corresponding model - - Cp_density_conductivity_data: Includes the Cp data and model, the density and the thermal conductivity data for the three resin systems under investigation: RTM6, 890RTM and XU3508 - -Thermal_conductivity_model: Includes the data from the thermal conductivity tests for the three resin system and the fitting of the experimental data with the analytical model. |
| Type Of Material | Database/Collection of data |
| Year Produced | 2018 |
| Provided To Others? | Yes |
| Impact | Dataset of thermal conductivity development on three industrial epoxy resin systems as well as a new analysis technique for obtaining these. |
| URL | https://figshare.com/articles/Measurement_of_thermal_conductivity_of_epoxy_resins_during_cure/584124... |
| Title | Multi-objective optimisation of Resin Infusion |
| Description | Dataset related to G Struzziero, AA Skordos. Multi-objective optimization of resin infusion. Advanced Manufacturing: Polymer & Composites Science. 2019;5:17-28 Exhaustive_analysis.xlsx: -'Exhaustive' sheet: All the data gathered during the exhaustive search are reported. Column 1 to 8 report respectively the temperature of first and second dwell, duration of first dwell and ramp, convection coefficient, gate location, filling time and maximum degree of cure. -'Gate-T1' sheet: Reports the data to plot the 3D landscape -'T1-T2' sheet: Reports the data to plot the 3D landscape Optimisation results.xlsx -'Results' sheet: Reports the 13 generations the GA requires to converge to a final optimal Pareto front. Column 1 to 9 report respectively the generation number, temperature of first and second dwell, duration of first dwell and ramp, gate location, convection coefficient, filling time and maximum degree of cure. -'Details of individuals' sheet: Reports the detailed analysis f the different strategies of solutions belonging to different region of the Pareto. Regions_analysis.xlsx -'Horizontal_flow_front_data' sheet: Reports the compact data for the Temperature, viscosity and degree of cure evolution for an horizontal region solution -'Viscosity_horizontal_region' sheet: Reports the viscosity evolution for different nodes in the model for an horizontal region solution -'Degree cure_horizontal_region' sheet: Reports the degree of cure envelope for an horizontal solution -'Temperature_horizontal region' sheet: Reports the temperature evolution for different nodes in the model for an horizontal region solution -'Model_verification' sheet: reports the analytical viscosity model implementation -'Vertical_flow front_data' sheet: Reports the compact data for the Temperature, viscosity and degree of cure evolution for a vertical region solution -'Degree_cure_vertical_region' sheet: Reports the degree of cure envelope for a vertical solution region -'Temperature_vertical_region' sheet: Reports the temperature evolution for different nodes in the model for a vertical region solution -'Viscosity_vertical_region' sheet: Reports the viscosity evolution for different nodes in the model for a vertical region solution -'Standard_flow_front_data' sheet: Reports the compact data for the Temperature, viscosity and degree of cure evolution for the standard solution -'Deg_cure_standard' sheet: Reports the degree of cure envelope for the standard solution '-Temperature_data_standard' sheet: Reports the temperature evolution for different nodes in the model for the standard solution -'Viscosity_standard' sheet: Reports the viscosity evolution for different nodes in the model for the standard solution -'Plots' sheet: plots summary |
| Type Of Material | Database/Collection of data |
| Year Produced | 2019 |
| Provided To Others? | Yes |
| Impact | Development of new infusion strategy based on optimised thermal profile to achieve efficient processing. |
| URL | https://doi.org/10.17862/cranfield.rd.5505376 |
| Title | Multi-objective optimisation of composites manufacturing under uncertainty |
| Description | Computational methodology for incorporation of variability in the objectives of process optimisation for filling and curing in composites manufacturing. The methodology involves and integration of a genetic algorithm, with a Kriging surrogate model of the process and Monte Carlo simulation. |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2017 |
| Provided To Others? | Yes |
| Impact | Incorporation of development in Clean Sky (H20202) developments through project SimCoDeQ. |
| Title | On line optimisation and active control of the cure process of thick composite laminates |
| Description | The dataset comprises experimental data, and modelling and simulation results reported in the paper titled 'On line optimisation and active control of the cure process of thick composite laminates' published in Journal of Manufacturing Processes (2023). The dataset is organised into 5 files: -'Kinetics-DSC.xlsx' contains the experimental DSC curves, after integration for the calculation of degree of cure evolution, and the corresponding cure kinetics modelling results. The file ia organised in two worksheets - 'Dynamic' reports the results of the one dynamic cure experiment and 'Isothermal' reports the results of the four isothermal experiments used in the analysis. -'Tg-MDSC.xlsx' contains the MDSC specific heat capacity measurements and the corresponding model results for three isothermal experiments in worksheet 'Specific heat capacity' and the glass transition temperature values measured for partially cured samples alongside the corresponding glass transistor temperature evolution model in worksheet 'Tg.' -'FE model validation.xlsx' contains the cure simulation validation trial results using a flat panel. The results are organised into two worksheets - 'Temperature' reports the experimental and simulation temperature as a function of time for 3 locations in the component and 'Degree of cure' the corresponding degree of cure results. -'Cable cure.xlsx' reports the results of the active control trial for the cure of a cable. Worksheet 'Active control trial' reports the setpoint evolution and the surface and centre temperatures measured, alongside the corresponding degree of cure and glass transition evolution. Worksheet 'Standard profile simulation' reports the simulation results for the nominal cure profile. -Multi objective optimisation.xlsx presents the Pareto (temperature overshoot versus cure time) front obtained from offline optimisation for a two-cure dwell profile using an exhaustive search alongside the values of cure time and temperature overshoot for the nominal cure profile and the actively controlled process. |
| Type Of Material | Database/Collection of data |
| Year Produced | 2023 |
| Provided To Others? | Yes |
| URL | https://cord.cranfield.ac.uk/articles/dataset/On_line_optimisation_and_active_control_of_the_cure_pr... |
| Title | On line optimisation and active control of the cure process of thick composite laminates |
| Description | The dataset comprises experimental data, and modelling and simulation results reported in the paper titled 'On line optimisation and active control of the cure process of thick composite laminates' published in Journal of Manufacturing Processes (2023). The dataset is organised into 5 files: -'Kinetics-DSC.xlsx' contains the experimental DSC curves, after integration for the calculation of degree of cure evolution, and the corresponding cure kinetics modelling results. The file ia organised in two worksheets - 'Dynamic' reports the results of the one dynamic cure experiment and 'Isothermal' reports the results of the four isothermal experiments used in the analysis. -'Tg-MDSC.xlsx' contains the MDSC specific heat capacity measurements and the corresponding model results for three isothermal experiments in worksheet 'Specific heat capacity' and the glass transition temperature values measured for partially cured samples alongside the corresponding glass transistor temperature evolution model in worksheet 'Tg.' -'FE model validation.xlsx' contains the cure simulation validation trial results using a flat panel. The results are organised into two worksheets - 'Temperature' reports the experimental and simulation temperature as a function of time for 3 locations in the component and 'Degree of cure' the corresponding degree of cure results. -'Cable cure.xlsx' reports the results of the active control trial for the cure of a cable. Worksheet 'Active control trial' reports the setpoint evolution and the surface and centre temperatures measured, alongside the corresponding degree of cure and glass transition evolution. Worksheet 'Standard profile simulation' reports the simulation results for the nominal cure profile. -Multi objective optimisation.xlsx presents the Pareto (temperature overshoot versus cure time) front obtained from offline optimisation for a two-cure dwell profile using an exhaustive search alongside the values of cure time and temperature overshoot for the nominal cure profile and the actively controlled process. |
| Type Of Material | Database/Collection of data |
| Year Produced | 2023 |
| Provided To Others? | Yes |
| URL | https://cord.cranfield.ac.uk/articles/dataset/On_line_optimisation_and_active_control_of_the_cure_pr... |
| Title | Real time inverse solution of the composites cure heat transfer problem under uncertainty |
| Description | -panel_glass.f : subroutines representing material properties and boundary conditions.-panel_glass_3.3mm.dat: Marc input corresponding to the cure model of this paper.-Process monitoring results.xlsx: Experimental data acquired by thermocouples.-Surrogate model validation.xlsx: Response surfaces of surrogate and FE model.-MCMC results: Statistical properties estimation of unknown parameters with inversion procedure.-Real time probability estimation: Minimum final degree of cure and minimum glass transition temperature probability estimation during manufacturing process. |
| Type Of Material | Database/Collection of data |
| Year Produced | 2019 |
| Provided To Others? | Yes |
| Impact | Validation of on line uncertainty estimation method applied to composite cure using temperature sensors. |
| URL | https://cord.cranfield.ac.uk/articles/Real_time_inverse_solution_of_the_composites_cure_heat_transfe... |
| Title | Real time uncertainty estimation in filling stage of RTM process |
| Description | Surrogate models validation: includes comparison between FE model and surrogate modelSensors data: includes the response of the three lineal sensorsReal time uncertainty estimation: includes the results of the inversion procedurePrior model: confidence intervals using prior knowledgePost model: confidence intervals using inversion solutionCDF Filling time estimation: Cumulative density function of filling time estimation at different times during filling process |
| Type Of Material | Database/Collection of data |
| Year Produced | 2020 |
| Provided To Others? | Yes |
| Impact | First application of on line inverse solution based flow monitoring data for estimation of process uncertainty. |
| URL | https://cord.cranfield.ac.uk/articles/dataset/Real_time_uncertainty_estimation_in_filling_stage_of_R... |
| Title | Surrogate process model |
| Description | Computationally efficient of composite manufacturing based on Kriging. This allows an off line execution of computationally intensive non linear process simulation producing results that are subsequently captured in the Kriging model. The later can be executed in real time as part of process control in a stochastic simulation/optimisation loop accompanied by on line process monitoring. The method has been demonstrated successfully for the curing stage of composites manufacturing, which constitutes the most computationally demanding simuation step. |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2017 |
| Provided To Others? | Yes |
| Impact | The method is a core element of a Clean Sky project SimCoDeq (Simulation tool development for a composite manufacturing process default prediction integrated into a quality control system) aiming at reducing variability in the manufacture of composite spars (Airbus Spain). |
| Title | Block Circulant Embedding Method |
| Description | The block circulant embedding method (BCEM) is a method for sampling from stationary Gaussian random field on a grid which is not regular but has a regular block structure which is often the case in applications. It was introduced in a paper titled "A Block Circulant Embedding Method for Simulation of Stationary Gaussian Random Fields on Block-regular Grids" by M. Park and M.V. Tretyakov. |
| Type Of Technology | Software |
| Year Produced | 2015 |
| Open Source License? | Yes |
| Impact | The introduced block-circulant embedding method (BCEM) can outperform the classical circulant embedding method (CEM), which requires a regularization of the irregular grid before its application. |
| URL | https://github.com/parkmh/bcempaper |