{"id":{"repo_id":"tu-berlin","oai_identifier":"oai:depositonce.tu-berlin.de:11303/22750"},"canonical_url":"https://search.dev.ndltd.org/etd/tu-berlin/oai:depositonce.tu-berlin.de:11303/22750","repository":{"repo_id":"tu-berlin","name":"Technische Universität Berlin","base_url":"https://api-depositonce.tu-berlin.de/server/oai/request"},"display":{"title":"Hybride Modellbildung zur Prozessführung von Fed-Batch-Kultivierungen mit Aspergillus niger","abstract":"Due to its high secretion capability, the filamentous growing fungus Aspergillus niger is a promising candidate for the industrial production of homologous and heterologous proteins. In this work, the growth and product formation of A.niger in fed-batch cultivations is modelled by using the enzyme glucoamylase as an example product. Since classical modeling of bioprocesses involves a great deal of effort, a hybrid modeling approach is being pursued that combines biologically motivated modeling with data-based methods and requires fewer biological details to be captured. In the first part, a data-based methodology for the rapid development of initial model structures is proposed, in which a pseudo-stoichiometry is estimated via singular value decomposition. Then, a linear transformation is used to convert the measured data into reaction-variant and -invariant states with which the kinetics of the model can be identified. The kinetics are described by a combination of a mechanistic model part and a multi-layer perceptron (MLP) as black box part. In the second part, a structured biological model for A.niger is derived, which also includes online measurement data and can therefore be used for process control. In this model, the product formation kinetics are replaced by an MLP, and different structures and strategies for training the MLP are investigated. A relationship between the glucose concentration in the reactor and the formation of glucoamylase can be revealed. The last part deals with an implementation of a sigma-point Kalman filter based on the derived model for the fed-batch process. In addition, a process optimization is performed to maximize the glucoamylase activity at the end of the cultivation.","abstract_html":"Due to its high secretion capability, the filamentous growing fungus Aspergillus niger is a promising candidate for the industrial production of homologous and heterologous proteins. In this work, the growth and product formation of A.niger in fed-batch cultivations is modelled by using the enzyme glucoamylase as an example product. Since classical modeling of bioprocesses involves a great deal of effort, a hybrid modeling approach is being pursued that combines biologically motivated modeling with data-based methods and requires fewer biological details to be captured. In the first part, a data-based methodology for the rapid development of initial model structures is proposed, in which a pseudo-stoichiometry is estimated via singular value decomposition. Then, a linear transformation is used to convert the measured data into reaction-variant and -invariant states with which the kinetics of the model can be identified. The kinetics are described by a combination of a mechanistic model part and a multi-layer perceptron (MLP) as black box part. In the second part, a structured biological model for A.niger is derived, which also includes online measurement data and can therefore be used for process control. In this model, the product formation kinetics are replaced by an MLP, and different structures and strategies for training the MLP are investigated. A relationship between the glucose concentration in the reactor and the formation of glucoamylase can be revealed. The last part deals with an implementation of a sigma-point Kalman filter based on the derived model for the fed-batch process. In addition, a process optimization is performed to maximize the glucoamylase activity at the end of the cultivation.","abstract_has_math":false,"creators":["Pößel, Till Alexander"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":["King, Rudibert"],"committee_chairs":[],"committee_members":[],"year":2024,"date_issued":"2024","date_published":"2024","updated_at":"2026-07-27T21:28:49Z","subjects":[],"languages":["de"],"rights":[],"rights_urls":["https://creativecommons.org/licenses/by/4.0/"],"identifier_entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://doi.org/10.14279/depositonce-21560"],"render_values":[{"text":"https://doi.org/10.14279/depositonce-21560","href":"https://doi.org/10.14279/depositonce-21560","code":true}]}]},"links":{"outbound_url":"https://depositonce.tu-berlin.de/handle/11303/22750","outbound_label":"Repository record","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["King, Rudibert"]},{"key":"dc:creator","label":"Author","values":["Pößel, Till Alexander"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2024-10-21T11:16:30Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2024-10-21T11:16:30Z"]},{"key":"dc:date.issued","label":"Date","values":["2024"]},{"key":"dc:type","label":"Dc Type","values":["Doctoral Thesis"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["de"]},{"key":"dc:rights.uri","label":"Rights URI","values":["https://creativecommons.org/licenses/by/4.0/"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://depositonce.tu-berlin.de/handle/11303/22750","https://doi.org/10.14279/depositonce-21560"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Due to its high secretion capability, the filamentous growing fungus Aspergillus niger is a promising candidate for the industrial production of homologous and heterologous proteins. In this work, the growth and product formation of A.niger in fed-batch cultivations is modelled by using the enzyme glucoamylase as an example product. Since classical modeling of bioprocesses involves a great deal of effort, a hybrid modeling approach is being pursued that combines biologically motivated modeling with data-based methods and requires fewer biological details to be captured. In the first part, a data-based methodology for the rapid development of initial model structures is proposed, in which a pseudo-stoichiometry is estimated via singular value decomposition. Then, a linear transformation is used to convert the measured data into reaction-variant and -invariant states with which the kinetics of the model can be identified. The kinetics are described by a combination of a mechanistic model part and a multi-layer perceptron (MLP) as black box part. In the second part, a structured biological model for A.niger is derived, which also includes online measurement data and can therefore be used for process control. In this model, the product formation kinetics are replaced by an MLP, and different structures and strategies for training the MLP are investigated. A relationship between the glucose concentration in the reactor and the formation of glucoamylase can be revealed. The last part deals with an implementation of a sigma-point Kalman filter based on the derived model for the fed-batch process. In addition, a process optimization is performed to maximize the glucoamylase activity at the end of the cultivation.","Der filamentös wachsende Pilz Aspergillus niger eignet sich aufgrund seiner Sekretionsfähigkeit zur industriellen Herstellung von homologen und heterologen Proteinen. Am Beispiel des Enzyms Glucoamylase wird in dieser Arbeit das Wachstums- und Produktbildungsverhalten von A.niger in Fed-Batch-Kultivierungen modelliert. Da die klassische Modellierung von Bioprozessen mit einem hohen Aufwand verbunden ist, wird ein hybrider Modellansatz verfolgt, der eine biologisch motivierte Modellbildung mit datenbasierten Methoden verknüpft und dabei weniger zu erfassende biologische Details benötigt. In einem ersten Schritt wird eine hauptsächlich datenbasierte Methodik zur schnellen Entwicklung von initialen Modellstrukturen vorgeschlagen, bei der die Stöchiometrie über eine Singulärwertzerlegung geschätzt wird. Über eine lineare Umformung werden anschließend die Messdaten in reaktionsvariante und -invariante Zustände überführt, mit denen die Kinetiken identifiziert werden können. Diese werden dabei durch eine Kombination von bekannten Modellanteilen und durch ein Multi-Layer Perceptron (MLP) als Black Box-Anteil beschrieben. Im zweiten Schritt wird ein strukturiertes, biologisches Modell für Fed-Batch-Kultivierungen von A.niger hergeleitet, das auch Online-Messgrößen umfasst und für die Prozessführung geeignet ist. In diesem Modell wird die Produktbildungskinetik von Glucoamylase durch ein MLP formuliert, wobei unterschiedliche Strukturen und Strategien für das Training des MLPs untersucht werden. Für die Bildung von Glucoamylase kann ein Zusammenhang mit der Glucosekonzentration im Reaktor aufgezeigt werden. Aufbauend auf dem hergeleiteten Modell wird anschließend ein Sigmapunkt-Kalman-Filter für den Fed-Batch-Prozess implementiert und zusätzlich eine Versuchsplanung mit dem Ziel der Maximierung der Glucoamylaseaktivität durchgeführt."]},{"key":"dc:title","label":"Title","values":["Hybride Modellbildung zur Prozessführung von Fed-Batch-Kultivierungen mit Aspergillus niger"]}]}],"canonical_facts":{"dc:contributor.advisor":["King, Rudibert"],"dc:creator":["Pößel, Till Alexander"],"dc:date.accessioned":["2024-10-21T11:16:30Z"],"dc:date.available":["2024-10-21T11:16:30Z"],"dc:date.issued":["2024"],"dc:description.abstract":["Due to its high secretion capability, the filamentous growing fungus Aspergillus niger is a promising candidate for the industrial production of homologous and heterologous proteins. In this work, the growth and product formation of A.niger in fed-batch cultivations is modelled by using the enzyme glucoamylase as an example product. Since classical modeling of bioprocesses involves a great deal of effort, a hybrid modeling approach is being pursued that combines biologically motivated modeling with data-based methods and requires fewer biological details to be captured. In the first part, a data-based methodology for the rapid development of initial model structures is proposed, in which a pseudo-stoichiometry is estimated via singular value decomposition. Then, a linear transformation is used to convert the measured data into reaction-variant and -invariant states with which the kinetics of the model can be identified. The kinetics are described by a combination of a mechanistic model part and a multi-layer perceptron (MLP) as black box part. In the second part, a structured biological model for A.niger is derived, which also includes online measurement data and can therefore be used for process control. In this model, the product formation kinetics are replaced by an MLP, and different structures and strategies for training the MLP are investigated. A relationship between the glucose concentration in the reactor and the formation of glucoamylase can be revealed. The last part deals with an implementation of a sigma-point Kalman filter based on the derived model for the fed-batch process. In addition, a process optimization is performed to maximize the glucoamylase activity at the end of the cultivation.","Der filamentös wachsende Pilz Aspergillus niger eignet sich aufgrund seiner Sekretionsfähigkeit zur industriellen Herstellung von homologen und heterologen Proteinen. Am Beispiel des Enzyms Glucoamylase wird in dieser Arbeit das Wachstums- und Produktbildungsverhalten von A.niger in Fed-Batch-Kultivierungen modelliert. Da die klassische Modellierung von Bioprozessen mit einem hohen Aufwand verbunden ist, wird ein hybrider Modellansatz verfolgt, der eine biologisch motivierte Modellbildung mit datenbasierten Methoden verknüpft und dabei weniger zu erfassende biologische Details benötigt. In einem ersten Schritt wird eine hauptsächlich datenbasierte Methodik zur schnellen Entwicklung von initialen Modellstrukturen vorgeschlagen, bei der die Stöchiometrie über eine Singulärwertzerlegung geschätzt wird. Über eine lineare Umformung werden anschließend die Messdaten in reaktionsvariante und -invariante Zustände überführt, mit denen die Kinetiken identifiziert werden können. Diese werden dabei durch eine Kombination von bekannten Modellanteilen und durch ein Multi-Layer Perceptron (MLP) als Black Box-Anteil beschrieben. Im zweiten Schritt wird ein strukturiertes, biologisches Modell für Fed-Batch-Kultivierungen von A.niger hergeleitet, das auch Online-Messgrößen umfasst und für die Prozessführung geeignet ist. In diesem Modell wird die Produktbildungskinetik von Glucoamylase durch ein MLP formuliert, wobei unterschiedliche Strukturen und Strategien für das Training des MLPs untersucht werden. Für die Bildung von Glucoamylase kann ein Zusammenhang mit der Glucosekonzentration im Reaktor aufgezeigt werden. Aufbauend auf dem hergeleiteten Modell wird anschließend ein Sigmapunkt-Kalman-Filter für den Fed-Batch-Prozess implementiert und zusätzlich eine Versuchsplanung mit dem Ziel der Maximierung der Glucoamylaseaktivität durchgeführt."],"dc:identifier.uri":["https://depositonce.tu-berlin.de/handle/11303/22750","https://doi.org/10.14279/depositonce-21560"],"dc:language.iso":["de"],"dc:rights.uri":["https://creativecommons.org/licenses/by/4.0/"],"dc:title":["Hybride Modellbildung zur Prozessführung von Fed-Batch-Kultivierungen mit Aspergillus niger"],"dc:type":["Doctoral Thesis"]},"updated_at":"2026-07-27T21:28:49Z"}