{"id":{"repo_id":"oxford-brookes","oai_identifier":"tle:5b0f53e3-9852-4e19-bbc6-2663ec5a05ca:d6bd9758-527a-46cd-bfe2-c433766e8fca:1"},"canonical_url":"https://search.dev.ndltd.org/etd/oxford-brookes/tle:5b0f53e3-9852-4e19-bbc6-2663ec5a05ca:d6bd9758-527a-46cd-bfe2-c433766e8fca:1","repository":{"repo_id":"oxford-brookes","name":"Oxford Brookes University","base_url":"https://radar.brookes.ac.uk/radar/oai"},"display":{"title":"CFD Modelling of Down-Sized GDI Engines: Fuel Injection, In-Cylinder Phenomena and Correlation to Engine-Out PN Emissions","abstract":"This PhD thesis aims to contribute improved knowledge and understanding of the particulate matter formation process in modern-design Gasoline Direct Injection engines. It does that through the development of a comprehensive 3D-CFD in-cylinder modelling tool that enables a robust and accurate insight into the pre-combustion mechanisms potentially leading to PM formation. The results presented refer to standard gasoline fuel, but remain fully relevant to synthetic hydrocarbons (efuels); the modelling approaches, on the other hand, are applicable to a wider range of modern and future fuels, including zero-carbon ones (particularly H2 and NH3). A 50-point engine testing matrix was designed and used as the basis to explore the impact of six relevant engine control variables on PN production. The randomised DoE approach enabled assessing a wide region of the typical passenger car operating envelope: engine load was varied between 20 and 140 Nm; engine speed in the range 2000 – 3500 rpm; injection pressure was varied between 60 and 250 bar; start of injection timing between 180 and 340 CAD BTCD; ignition timing was varied between 0 and 40 CAD BTDC, and finally intake valve opening timing in the range 0 – 40 CAD BTDCi. An extensive CFD modelling campaign was conducted based on this test matrix, and parallel experimental tests used to measure cylinder-out PN and to collect data as simulation input and for the purpose of calibrating and/or validating the models. The CFD model was developed targeting, wherever possible, ease of implementation and industrial utilisation, using approaches, sub-models and data that should be available to large corporations in the automotive sector. The whole model was developed using commercial CFD software Siemens Star-CD version 2021.1-436008. Fuel injection was modelled using a combination of initial Rosin-Rammler droplet size distribution and the Reitz-Diwaker secondary break-up model. An initial statistical variable screening methodology was followed to minimise the number of calibration factors, reducing them from six to three. These were the coefficients X and q from the Rosin-Rammler distribution and the ‘Bag break-up’ coefficient for the Reitz-Diwaker model. Statistical and optimisation techniques enable spray model calibration at four representative levels of injection pressure: 35, 100, 150 and 200 bar. The results demonstrated good agreement in terms of spray tip penetration, average droplet diameter (SMD) and overall spray morphology. A Ricardo WAVE engine model, made available by Ford Motor Company, was repurposed and calibrated to match experimental in-cylinder trapped mass within 2% error, across the operating space, using valve lash and pressure drop across the throttle; this model was then used to provide CA-resolved intake and exhaust pressure and temperature boundary conditions for the CFD simulations. In-cylinder surface temperatures were proven by this study to be an essential input to accurately predict the dynamics of liquid film formation and, in turn, of gas-phase mixture quality. These inputs were defined based on speed-load correlations derived from piston, liner and cylinder head telemetry data. The process of liquid spray-to-wall impact, subsequent liquid film deposition and film-to-flow evaporation were modelled via the Bai-Onera approach and Habchi Leidenfrost temperature definition. The validated CFD model was used to aid in exploring pre-combustion metrics that may potentially enable PN production. Metrics extracted from CFD included the mixture Uniformity Index of Phi, 〖UI〗_φ, the mass of fuel in richer-than-expected regions, 〖M+〗_φ, and various liquid film quantities. As a general rule, low mixture homogeneity at spark timing and/or high levels of retained liquid film led to high cylinder-out PN, and vice versa. As an example of good quality mixture preparation, OP47 at 2632 rpm and 23 Nm, with an injection pressure of 215 bar and SOI of 331 CAD BTDCc, 〖UI〗_φ was 0.937, 〖M+〗_1.1 was 1.105 mg and LF was 0.021 mg; correspondingly, the measured PN was 7.09 x106 #/cc. By contrast, as an example of poor quality, at 2888 rpm and 96 Nm, with an injection pressure of 62 bar and SOI of 226 CAD BTDCc, 〖UI〗_φ was 0.838, 〖M+〗_1.1 was 12.253 mg and LF was 1.418 mg; correspondingly, the measured PN was 2.60 x108 #/cc. While operating points with either clearly good or clearly bad pre-combustion metrics were generally easier to categorise in terms of their PN yield, a limited number of mid-level PN conditions appeared to elude any possible categorisation. This pointed to potential limitations of the approach including accuracy of the PN measurements or, more likely, accuracy of the wall temperature measurements – both being inherently difficult to acquire. Supplementary in-cylinder simulations were then carried out to explore the impact of surface temperature on liquid film retained on top of the piston, which has been demonstrated to lead invariably to pool fire and high PN output. This work, which culminates with the definition of the concept of ‘Limiting Temperature’ for the retainment of liquid film, has significant potential consequences for the identification of control strategies and feasible metal surface modifications to minimise the sooting tendency in modern GDI engines, and for the development of new synthetic fuels with prescribed thermo-physical properties. Finally, an elastic net regression methodology is used to generate simple surrogate models linking cylinder-out PN and engine control parameters or supplementary variables derived from those. This study is deemed as a useful addition, as it reveals powerful correlations that may be considered during engine calibration work, when the aid of complex CFD modelling is not an option. The best and, at the same time, the easiest EN model uses only two critical metrics, injection pulse width and available mixing time (time elapsed between EOI and Spark Timing), to predict PN to within 5% accuracy.","abstract_html":"This PhD thesis aims to contribute improved knowledge and understanding of the particulate matter formation process in modern-design Gasoline Direct Injection engines. It does that through the development of a comprehensive 3D-CFD in-cylinder modelling tool that enables a robust and accurate insight into the pre-combustion mechanisms potentially leading to PM formation. The results presented refer to standard gasoline fuel, but remain fully relevant to synthetic hydrocarbons (efuels); the modelling approaches, on the other hand, are applicable to a wider range of modern and future fuels, including zero-carbon ones (particularly H2 and NH3). A 50-point engine testing matrix was designed and used as the basis to explore the impact of six relevant engine control variables on PN production. The randomised DoE approach enabled assessing a wide region of the typical passenger car operating envelope: engine load was varied between 20 and 140 Nm; engine speed in the range 2000 – 3500 rpm; injection pressure was varied between 60 and 250 bar; start of injection timing between 180 and 340 CAD BTCD; ignition timing was varied between 0 and 40 CAD BTDC, and finally intake valve opening timing in the range 0 – 40 CAD BTDCi. An extensive CFD modelling campaign was conducted based on this test matrix, and parallel experimental tests used to measure cylinder-out PN and to collect data as simulation input and for the purpose of calibrating and/or validating the models. The CFD model was developed targeting, wherever possible, ease of implementation and industrial utilisation, using approaches, sub-models and data that should be available to large corporations in the automotive sector. The whole model was developed using commercial CFD software Siemens Star-CD version 2021.1-436008. Fuel injection was modelled using a combination of initial Rosin-Rammler droplet size distribution and the Reitz-Diwaker secondary break-up model. An initial statistical variable screening methodology was followed to minimise the number of calibration factors, reducing them from six to three. These were the coefficients X and q from the Rosin-Rammler distribution and the ‘Bag break-up’ coefficient for the Reitz-Diwaker model. Statistical and optimisation techniques enable spray model calibration at four representative levels of injection pressure: 35, 100, 150 and 200 bar. The results demonstrated good agreement in terms of spray tip penetration, average droplet diameter (SMD) and overall spray morphology. A Ricardo WAVE engine model, made available by Ford Motor Company, was repurposed and calibrated to match experimental in-cylinder trapped mass within 2% error, across the operating space, using valve lash and pressure drop across the throttle; this model was then used to provide CA-resolved intake and exhaust pressure and temperature boundary conditions for the CFD simulations. In-cylinder surface temperatures were proven by this study to be an essential input to accurately predict the dynamics of liquid film formation and, in turn, of gas-phase mixture quality. These inputs were defined based on speed-load correlations derived from piston, liner and cylinder head telemetry data. The process of liquid spray-to-wall impact, subsequent liquid film deposition and film-to-flow evaporation were modelled via the Bai-Onera approach and Habchi Leidenfrost temperature definition. The validated CFD model was used to aid in exploring pre-combustion metrics that may potentially enable PN production. Metrics extracted from CFD included the mixture Uniformity Index of Phi, 〖UI〗_φ, the mass of fuel in richer-than-expected regions, 〖M+〗_φ, and various liquid film quantities. As a general rule, low mixture homogeneity at spark timing and/or high levels of retained liquid film led to high cylinder-out PN, and vice versa. As an example of good quality mixture preparation, OP47 at 2632 rpm and 23 Nm, with an injection pressure of 215 bar and SOI of 331 CAD BTDCc, 〖UI〗_φ was 0.937, 〖M+〗_1.1 was 1.105 mg and LF was 0.021 mg; correspondingly, the measured PN was 7.09 x106 #/cc. By contrast, as an example of poor quality, at 2888 rpm and 96 Nm, with an injection pressure of 62 bar and SOI of 226 CAD BTDCc, 〖UI〗_φ was 0.838, 〖M+〗_1.1 was 12.253 mg and LF was 1.418 mg; correspondingly, the measured PN was 2.60 x108 #/cc. While operating points with either clearly good or clearly bad pre-combustion metrics were generally easier to categorise in terms of their PN yield, a limited number of mid-level PN conditions appeared to elude any possible categorisation. This pointed to potential limitations of the approach including accuracy of the PN measurements or, more likely, accuracy of the wall temperature measurements – both being inherently difficult to acquire. Supplementary in-cylinder simulations were then carried out to explore the impact of surface temperature on liquid film retained on top of the piston, which has been demonstrated to lead invariably to pool fire and high PN output. This work, which culminates with the definition of the concept of ‘Limiting Temperature’ for the retainment of liquid film, has significant potential consequences for the identification of control strategies and feasible metal surface modifications to minimise the sooting tendency in modern GDI engines, and for the development of new synthetic fuels with prescribed thermo-physical properties. Finally, an elastic net regression methodology is used to generate simple surrogate models linking cylinder-out PN and engine control parameters or supplementary variables derived from those. This study is deemed as a useful addition, as it reveals powerful correlations that may be considered during engine calibration work, when the aid of complex CFD modelling is not an option. The best and, at the same time, the easiest EN model uses only two critical metrics, injection pulse width and available mixing time (time elapsed between EOI and Spark Timing), to predict PN to within 5% accuracy.","abstract_has_math":false,"creators":["Hopkins, Edward"],"institution":"Oxford Brookes University","degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":["Bonatesta, Fabrizio","Bell, Daniel"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":null,"date_issued":"","date_published":null,"updated_at":"2026-07-24T03:42:15Z","subjects":[],"languages":["en"],"rights":["All rights reserved"],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://doi.org/10.24384/5a1b-9b39","outbound_label":"DOI","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Bonatesta, Fabrizio","Bell, Daniel","Hopkins, Edward"]},{"key":"dc:creator","label":"Author","values":["Hopkins, Edward"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:publisher","label":"Institution","values":["Oxford Brookes University"]},{"key":"dc:type","label":"Dc Type","values":["thesis"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["en"]},{"key":"dc:rights","label":"Dc Rights","values":["All rights reserved"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://doi.org/10.24384/5a1b-9b39","https://radar.brookes.ac.uk/radar/file/5b0f53e3-9852-4e19-bbc6-2663ec5a05ca/1/Hopkins2024GDIEngines.pdf"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["This PhD thesis aims to contribute improved knowledge and understanding of the particulate matter formation process in modern-design Gasoline Direct Injection engines. It does that through the development of a comprehensive 3D-CFD in-cylinder modelling tool that enables a robust and accurate insight into the pre-combustion mechanisms potentially leading to PM formation. The results presented refer to standard gasoline fuel, but remain fully relevant to synthetic hydrocarbons (efuels); the modelling approaches, on the other hand, are applicable to a wider range of modern and future fuels, including zero-carbon ones (particularly H2 and NH3). A 50-point engine testing matrix was designed and used as the basis to explore the impact of six relevant engine control variables on PN production. The randomised DoE approach enabled assessing a wide region of the typical passenger car operating envelope: engine load was varied between 20 and 140 Nm; engine speed in the range 2000 – 3500 rpm; injection pressure was varied between 60 and 250 bar; start of injection timing between 180 and 340 CAD BTCD; ignition timing was varied between 0 and 40 CAD BTDC, and finally intake valve opening timing in the range 0 – 40 CAD BTDCi. An extensive CFD modelling campaign was conducted based on this test matrix, and parallel experimental tests used to measure cylinder-out PN and to collect data as simulation input and for the purpose of calibrating and/or validating the models. The CFD model was developed targeting, wherever possible, ease of implementation and industrial utilisation, using approaches, sub-models and data that should be available to large corporations in the automotive sector. The whole model was developed using commercial CFD software Siemens Star-CD version 2021.1-436008. Fuel injection was modelled using a combination of initial Rosin-Rammler droplet size distribution and the Reitz-Diwaker secondary break-up model. An initial statistical variable screening methodology was followed to minimise the number of calibration factors, reducing them from six to three. These were the coefficients X and q from the Rosin-Rammler distribution and the ‘Bag break-up’ coefficient for the Reitz-Diwaker model. Statistical and optimisation techniques enable spray model calibration at four representative levels of injection pressure: 35, 100, 150 and 200 bar. The results demonstrated good agreement in terms of spray tip penetration, average droplet diameter (SMD) and overall spray morphology. A Ricardo WAVE engine model, made available by Ford Motor Company, was repurposed and calibrated to match experimental in-cylinder trapped mass within 2% error, across the operating space, using valve lash and pressure drop across the throttle; this model was then used to provide CA-resolved intake and exhaust pressure and temperature boundary conditions for the CFD simulations. In-cylinder surface temperatures were proven by this study to be an essential input to accurately predict the dynamics of liquid film formation and, in turn, of gas-phase mixture quality. These inputs were defined based on speed-load correlations derived from piston, liner and cylinder head telemetry data. The process of liquid spray-to-wall impact, subsequent liquid film deposition and film-to-flow evaporation were modelled via the Bai-Onera approach and Habchi Leidenfrost temperature definition. The validated CFD model was used to aid in exploring pre-combustion metrics that may potentially enable PN production. Metrics extracted from CFD included the mixture Uniformity Index of Phi, 〖UI〗_φ, the mass of fuel in richer-than-expected regions, 〖M+〗_φ, and various liquid film quantities. As a general rule, low mixture homogeneity at spark timing and/or high levels of retained liquid film led to high cylinder-out PN, and vice versa. As an example of good quality mixture preparation, OP47 at 2632 rpm and 23 Nm, with an injection pressure of 215 bar and SOI of 331 CAD BTDCc, 〖UI〗_φ was 0.937, 〖M+〗_1.1 was 1.105 mg and LF was 0.021 mg; correspondingly, the measured PN was 7.09 x106 #/cc. By contrast, as an example of poor quality, at 2888 rpm and 96 Nm, with an injection pressure of 62 bar and SOI of 226 CAD BTDCc, 〖UI〗_φ was 0.838, 〖M+〗_1.1 was 12.253 mg and LF was 1.418 mg; correspondingly, the measured PN was 2.60 x108 #/cc. While operating points with either clearly good or clearly bad pre-combustion metrics were generally easier to categorise in terms of their PN yield, a limited number of mid-level PN conditions appeared to elude any possible categorisation. This pointed to potential limitations of the approach including accuracy of the PN measurements or, more likely, accuracy of the wall temperature measurements – both being inherently difficult to acquire. Supplementary in-cylinder simulations were then carried out to explore the impact of surface temperature on liquid film retained on top of the piston, which has been demonstrated to lead invariably to pool fire and high PN output. This work, which culminates with the definition of the concept of ‘Limiting Temperature’ for the retainment of liquid film, has significant potential consequences for the identification of control strategies and feasible metal surface modifications to minimise the sooting tendency in modern GDI engines, and for the development of new synthetic fuels with prescribed thermo-physical properties. Finally, an elastic net regression methodology is used to generate simple surrogate models linking cylinder-out PN and engine control parameters or supplementary variables derived from those. This study is deemed as a useful addition, as it reveals powerful correlations that may be considered during engine calibration work, when the aid of complex CFD modelling is not an option. The best and, at the same time, the easiest EN model uses only two critical metrics, injection pulse width and available mixing time (time elapsed between EOI and Spark Timing), to predict PN to within 5% accuracy."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["CFD Modelling of Down-Sized GDI Engines: Fuel Injection, In-Cylinder Phenomena and Correlation to Engine-Out PN Emissions"]}]}],"canonical_facts":{"dc:contributor":["Bonatesta, Fabrizio","Bell, Daniel","Hopkins, Edward"],"dc:creator":["Hopkins, Edward"],"dc:description":["This PhD thesis aims to contribute improved knowledge and understanding of the particulate matter formation process in modern-design Gasoline Direct Injection engines. It does that through the development of a comprehensive 3D-CFD in-cylinder modelling tool that enables a robust and accurate insight into the pre-combustion mechanisms potentially leading to PM formation. The results presented refer to standard gasoline fuel, but remain fully relevant to synthetic hydrocarbons (efuels); the modelling approaches, on the other hand, are applicable to a wider range of modern and future fuels, including zero-carbon ones (particularly H2 and NH3). A 50-point engine testing matrix was designed and used as the basis to explore the impact of six relevant engine control variables on PN production. The randomised DoE approach enabled assessing a wide region of the typical passenger car operating envelope: engine load was varied between 20 and 140 Nm; engine speed in the range 2000 – 3500 rpm; injection pressure was varied between 60 and 250 bar; start of injection timing between 180 and 340 CAD BTCD; ignition timing was varied between 0 and 40 CAD BTDC, and finally intake valve opening timing in the range 0 – 40 CAD BTDCi. An extensive CFD modelling campaign was conducted based on this test matrix, and parallel experimental tests used to measure cylinder-out PN and to collect data as simulation input and for the purpose of calibrating and/or validating the models. The CFD model was developed targeting, wherever possible, ease of implementation and industrial utilisation, using approaches, sub-models and data that should be available to large corporations in the automotive sector. The whole model was developed using commercial CFD software Siemens Star-CD version 2021.1-436008. Fuel injection was modelled using a combination of initial Rosin-Rammler droplet size distribution and the Reitz-Diwaker secondary break-up model. An initial statistical variable screening methodology was followed to minimise the number of calibration factors, reducing them from six to three. These were the coefficients X and q from the Rosin-Rammler distribution and the ‘Bag break-up’ coefficient for the Reitz-Diwaker model. Statistical and optimisation techniques enable spray model calibration at four representative levels of injection pressure: 35, 100, 150 and 200 bar. The results demonstrated good agreement in terms of spray tip penetration, average droplet diameter (SMD) and overall spray morphology. A Ricardo WAVE engine model, made available by Ford Motor Company, was repurposed and calibrated to match experimental in-cylinder trapped mass within 2% error, across the operating space, using valve lash and pressure drop across the throttle; this model was then used to provide CA-resolved intake and exhaust pressure and temperature boundary conditions for the CFD simulations. In-cylinder surface temperatures were proven by this study to be an essential input to accurately predict the dynamics of liquid film formation and, in turn, of gas-phase mixture quality. These inputs were defined based on speed-load correlations derived from piston, liner and cylinder head telemetry data. The process of liquid spray-to-wall impact, subsequent liquid film deposition and film-to-flow evaporation were modelled via the Bai-Onera approach and Habchi Leidenfrost temperature definition. The validated CFD model was used to aid in exploring pre-combustion metrics that may potentially enable PN production. Metrics extracted from CFD included the mixture Uniformity Index of Phi, 〖UI〗_φ, the mass of fuel in richer-than-expected regions, 〖M+〗_φ, and various liquid film quantities. As a general rule, low mixture homogeneity at spark timing and/or high levels of retained liquid film led to high cylinder-out PN, and vice versa. As an example of good quality mixture preparation, OP47 at 2632 rpm and 23 Nm, with an injection pressure of 215 bar and SOI of 331 CAD BTDCc, 〖UI〗_φ was 0.937, 〖M+〗_1.1 was 1.105 mg and LF was 0.021 mg; correspondingly, the measured PN was 7.09 x106 #/cc. By contrast, as an example of poor quality, at 2888 rpm and 96 Nm, with an injection pressure of 62 bar and SOI of 226 CAD BTDCc, 〖UI〗_φ was 0.838, 〖M+〗_1.1 was 12.253 mg and LF was 1.418 mg; correspondingly, the measured PN was 2.60 x108 #/cc. While operating points with either clearly good or clearly bad pre-combustion metrics were generally easier to categorise in terms of their PN yield, a limited number of mid-level PN conditions appeared to elude any possible categorisation. This pointed to potential limitations of the approach including accuracy of the PN measurements or, more likely, accuracy of the wall temperature measurements – both being inherently difficult to acquire. Supplementary in-cylinder simulations were then carried out to explore the impact of surface temperature on liquid film retained on top of the piston, which has been demonstrated to lead invariably to pool fire and high PN output. This work, which culminates with the definition of the concept of ‘Limiting Temperature’ for the retainment of liquid film, has significant potential consequences for the identification of control strategies and feasible metal surface modifications to minimise the sooting tendency in modern GDI engines, and for the development of new synthetic fuels with prescribed thermo-physical properties. Finally, an elastic net regression methodology is used to generate simple surrogate models linking cylinder-out PN and engine control parameters or supplementary variables derived from those. This study is deemed as a useful addition, as it reveals powerful correlations that may be considered during engine calibration work, when the aid of complex CFD modelling is not an option. The best and, at the same time, the easiest EN model uses only two critical metrics, injection pulse width and available mixing time (time elapsed between EOI and Spark Timing), to predict PN to within 5% accuracy."],"dc:format":["application/pdf"],"dc:identifier":["https://doi.org/10.24384/5a1b-9b39","https://radar.brookes.ac.uk/radar/file/5b0f53e3-9852-4e19-bbc6-2663ec5a05ca/1/Hopkins2024GDIEngines.pdf"],"dc:language":["en"],"dc:publisher":["Oxford Brookes University"],"dc:rights":["All rights reserved"],"dc:title":["CFD Modelling of Down-Sized GDI Engines: Fuel Injection, In-Cylinder Phenomena and Correlation to Engine-Out PN Emissions"],"dc:type":["thesis"]},"updated_at":"2026-07-24T03:42:15Z"}