WO2019086610A1 - Procédé de fabrication, procédé de qualification et/ou de mise en service d'une fabrication industrielle, et qualification et mise en service d'équipement et de processus - Google Patents

Procédé de fabrication, procédé de qualification et/ou de mise en service d'une fabrication industrielle, et qualification et mise en service d'équipement et de processus Download PDF

Info

Publication number
WO2019086610A1
WO2019086610A1 PCT/EP2018/080018 EP2018080018W WO2019086610A1 WO 2019086610 A1 WO2019086610 A1 WO 2019086610A1 EP 2018080018 W EP2018080018 W EP 2018080018W WO 2019086610 A1 WO2019086610 A1 WO 2019086610A1
Authority
WO
WIPO (PCT)
Prior art keywords
requirements
manufacturing
medical devices
medicinal products
labelling
Prior art date
Application number
PCT/EP2018/080018
Other languages
English (en)
Inventor
Per Rønsholt Andresen
Original Assignee
Pharma4Ever A/S
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Pharma4Ever A/S filed Critical Pharma4Ever A/S
Publication of WO2019086610A1 publication Critical patent/WO2019086610A1/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
    • G06Q10/087—Inventory or stock management, e.g. order filling, procurement or balancing against orders
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Definitions

  • the present disclosure relates to a manufacturing process and system for medicinal products or medical devices and to a method for qualifying and commissioning an industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities related to said process.
  • the disclosure further relates to a computer-implemented requirement generation system for generating a set of requirements for said process and/or equipment or facilities.
  • GMP Good Manufacturing Practices
  • qualification commissioning and validation of manufacturing, labelling, processing, assembling and packaging of medicinal products
  • manufacturers of medicinal products are required to qualify facilities, equipment, utilities and processes used for the manufacturing of medicinal products.
  • the EudraLex, Volume 4, EU Guidelines for Good Manufacturing Practice for Medicinal Products for Human and Veterinary Use, Annex 15: Qualification and Validation describes the principles of qualification and validation which are applicable to the facilities, equipment, utilities and processes used for the manufacture of medicinal products. It is set out that qualification, commissioning and validation should be based on justified and documented risk assessment of the facilities, equipment, utilities and processes. Quality risk management approach should be applied throughout the lifecycle of a medicinal product. Retrospective validation is not considered an acceptable approach.
  • requirements requirements, detailed technical implementation requirements, as well as qualification of installation, operation and performance, needs to be generated managed and actively used in manufacturing, processing, labelling, assembling and/or packaging process of medicinal products.
  • the inter-relationship between requirements should be clearly defined and validation protocols should define the critical systems, attributes and parameters and associated acceptance criteria. Challenges in these systems include that it is a very time-consuming and sometimes complex task to collect all requirements, correctly identify relevant and valid dependencies between the qualification stages, apply the requirements to the manufacturing, labelling, assembling and/or packaging process properly, and maintain and update the complete set of requirements for all qualification stages throughout the lifecycle of the product and process. Moreover, if requirements in some of the qualification stages are changed during the lifecycle, requirements of other qualification stages may have to be updated accordingly, which may also be a complex and time-consuming task.
  • the present disclosure relates to a method for qualifying and/or commissioning an industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities related to said process, to a manufacturing process for medicinal products or medical devices, to manufacturing control system and to a computer- implemented requirement generation system for generating a set of requirements for an industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities used in said manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices.
  • the method and systems may be used for improved quality of manufacturing processes and for automation of the validation lifecycle of manufacturing of pharmaceutical and biotechnology products.
  • ANNs Artificial neural networks
  • Neural networks may be seen as machine learning models employing model layers to predict an output for a received input.
  • Recurrent neural networks are a class of ANNs.
  • RNNs have an internal memory in order to process sequential inputs.
  • an RNN includes an input layer, a hidden layer, and an output layer.
  • the output depends on both the input and the previous state of the network when sequential inputs are processed. RNNs are therefore applied to areas such as writing and speech recognition.
  • a recursive neural network is a kind of deep neural network created by applying the same set of weights recursively over a structure, to produce a structured prediction over variable-size input structures, or a scalar prediction on it, by traversing a given structure in topological order.
  • a convolutional neural network (CNN) is a class of deep, feed-forward artificial neural networks, most commonly applied to analyzing visual imagery. 'Convolutional' refers to convolutional layers applying a convolution operation of the input, passing the result to the next layer.
  • a first embodiment of the presently disclosed manufacturing process for medicinal products or medical devices comprises the steps of: obtaining a first set of
  • the inventor has realized that two recurrent networks arranged in a sequence-to-sequence configuration provide an efficient way of handling requirements related to technical characteristic of the manufacturing process.
  • An embodiment of the presently disclosed computer-implemented requirement generation system for generating a set of requirements for an industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities used in said manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices comprises at least one hardware processor and at least one storage device storing requirement generation instructions and a requirement generation model, wherein the requirement generation model comprises:
  • the requirement generation instructions when executed by the at least one hardware processor, causes the requirement generation system to generate the second set of requirements based on an update of or the provision of a new first set of requirements.
  • the requirement generation model may thereby comprise a plurality of sets of requirements relating to a use, a technical specification or a technical qualification of the industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities and a recursive neural network model trained to translate a first set of requirements to a second set of requirements.
  • the inventors have, surprisingly, found that if requirements related to industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or equipment or facilities used in said manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products are arranged in a sequence and two recurrent networks are arranged in a sequence-to-sequence configuration, wherein there is an internal agreement of format providing an encoding of a first set of requirement to an intermediate set of characters, preferably in predefined format, and a subsequent decoding of the intermediate set of characters to a second set of requirements, then the system is capable of converting different types of requirements relating to for example user requirements, functional requirement specifications for the equipment, facilities and/or utilities or the like, and installation qualification, operational qualification and performance qualification.
  • the system enables a powerful and efficient
  • the system can be used both continuously in a manufacturing, formulation and filling, labelling, processing, assembling and/or packaging system when the conditions change for example in that some requirements are updated or by adding new requirements.
  • the requirements may be divided into different groups or categories, for example according to a V-model, wherein an industrial process related to for example manufacturing of pharmaceuticals is managed by a specification phase and a qualification and validation phase.
  • the first set and second set of requirements may thereby be transformed from any group to any other group according to the presently disclosed method.
  • the present disclosure relates to a method for qualifying and/or commissioning an industrial manufacturing, formulation and filling, labelling, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities related to said process, said method comprising the steps of:
  • the above method for qualifying and/or commissioning an industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities related to said process may be carried out by any embodiment of the presently disclosed computer-implemented requirement generation system.
  • the 'requirement generation system' is thereby not limited to a system only for requirement generation system but may additionally be used a system for controlling and/or verifying the industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities related to said process.
  • Said qualification and/or commissioning method may be applied on all equipment covered by the Good Manufacturing practices for Medicinal Products for Human and Veterinary Use. Said method can be seen as a way of reducing a complex and time- consuming task (qualification of an industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or equipment or facilities related to said process) to a more automatic and systematic approach based on confirmed data and operation by grouping and using requirements as sequential data and applying two recurrent neural networks in a sequence-to-sequence configuration.
  • Fig. 1 shows an embodiment of the presently disclosed method for qualifying and/or commissioning a process and/or equipment.
  • Fig. 2 shows a schematic illustration of an embodiment of the presently disclosed computer-implemented requirement generation system for generating a set of requirements for an industrial process and/or equipment.
  • Fig. 3 shows an example of a setup, wherein two recurrent neural networks are arranged in a sequence-to-sequence configuration for operating according to the presently disclosed computer-implemented requirement generation system and qualification/commissioning method.
  • Fig. 4 shows an example of groups of requirements according to a V-model.
  • Fig. 5 shows an illustration of how different types of requirements may influence each other.
  • Fig. 6 shows an embodiment of the presently disclosed manufacturing process for medicinal products or medical devices.
  • Fig. 7 shows a schematic illustration of an embodiment of the presently disclosed manufacturing control system.
  • 'Requirement' shall be construed broadly as any requirement from a user, from a supplier of equipment, an authority, a technical requirement on the equipment itself, a functional requirement, a specification requirement, but also as a test requirement or a verification or validation requirement defining how a test, verification or validation has to be carried out and the parameters that have to be verified accordingly.
  • 'Medical device' shall be construed broadly as any apparatus, appliance, software, material or other article, whether used alone or combination, intended by the manufacturer to be used for diagnosis, prevention, monitoring, treatment, including palliative and/or ameliorative treatment, or alleviation of disease, injury, handicap or other physiological process.
  • 'Medicinal product' shall be construed as any substance or combination of substances having, or presented as having, treating or preventing disease for human and/or veterinary use, or which may be used in or administered for human and/or veterinary use either with a view to restoring, correcting or modifying physiological functions by exerting a pharmacological, immunological or metabolic action, or to making a medical diagnosis, and for palliative and/or ameliorative treatment, or alleviation.
  • 'Medicinal product' may thereby also comprise products derived from blood or plasma.
  • Any use of 'medicinal products and/or medical devices' in the present disclosure shall be construed to comprise combinations of medicinal products and medical devices, such as medical devices for injecting a substance.
  • 'Security' in the context of the present invention refers to any safety-related parameter of a manufacturing, formulation and filling, processing, labelling, assembling or packaging process or related equipment.
  • the term is given the meaning that a requirement may comprise information related to 'security' in the sense that one or several parameters may be defined to express how a process needs to be carried out or what the equipment needs to fulfil to be considered safe. Examples of aspects that may be included in the term are requirements that a manufactured product may not be damaged, or heated above a threshold, or shaken beyond a threshold.
  • 'Qualification' in the context of the present invention refers to the process of determining whether facilities, equipment, utilities and processes relating to
  • Commissioning' in the context of the present invention refers to the process of finding potential errors or risks in facilities, equipment, utilities and processes relating to manufacturing, formulation and filling, labelling, processing, assembling and/or packaging of medicinal products and/or medical devices.
  • a commissioning process typically also involves correction or improvement of errors in the system.
  • the present disclosure relates to a method for qualifying and commissioning an industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities related to said process.
  • the disclosure further relates to a computer-implemented requirement generation system for generating a set of requirements for said process and/or equipment or facilities.
  • the inventor has found that by applying an artificial neural network to a context wherein often large amount of requirements related to the qualification of equipment and processes, a significantly more efficient and reliable handling can be achieved.
  • the requirements could come from the manufacturer of medicinal products and/or medical devices, from the suppliers of software, hardware and/or machines and/or from authorities and have a high degree of dependency between each other.
  • dependencies may also be influenced by interpretation, including different understanding of what a specific requirement means. Because of the highly complex environment, and, to some extent, subjective dependencies, it has turned out to be a difficult, if not impossible task, to use a clear set of rules for handling and generating requirements for industrial
  • the system may be further operative to control, or at least contribute in the control of the industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities used in said manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices based on the generated requirements.
  • the presently disclosed manufacturing process for medicinal products or medical devices may comprise a first step of: obtaining a first set of requirements related to first technical characteristics of said manufacturing of medicinal products or medical devices.
  • the method may further comprise a second step of obtaining a second set of requirements related to second technical characteristics of said manufacturing of medicinal products or medical devices.
  • the first set of requirements may for example related to security of the manufacturing process, but could also be related to how the medicinal products or medical devices are produced, transported, gripped or otherwise physically handled, such as within which pressure and/or temperature and/or moisture ranges that the medicinal products or medical devices are handled, in the
  • the first set of requirements may be selected from the group of user requirement specifications, identified risks and/or hazards, functional design specifications and detailed design specification.
  • the second set of requirements may, accordingly, be selected from the group of installation qualification, operational qualification and performance
  • the second set of requirements may, for example, be related to correct identification of defect manufactured products and/or whether equipment used in the manufacturing process has a predefined minimum level of stability for a given intensity of operation and/or whether the product has been produced within correct physical limits, such as pressure and/or temperature.
  • the first set of requirements may comprise requirements related to technical verification of the manufacturing process
  • the second set of requirements may comprise requirements related to technical validation of the manufacturing process. The technical verification and technical validation may thereby correspond to a v-model of the manufacturing process.
  • the method may further comprise the steps of sequentially encode the first set of requirements to an encoded intermediate vector of numbers or characters using a first recurrent neural network model; and sequentially decode the encoded intermediate vector of numbers or characters to the second set of requirements using a second recurrent neural network model.
  • the manufacturing of the medicinal products or medical devices may thereby take into account at least the second set of requirements; and implicitly also the first set of requirements, which have been used to generate the second set of requirements.
  • the method may further comprise the step of validating that requirements, such as the second set of requirements are fulfilled in the step of manufacturing the medicinal products or medical devices.
  • the skilled person will appreciate that the manufacturing process could also, by analogy, be a formulation and filling, labelling, assembling or packing process for medicinal products or medical devices.
  • the present disclosure further relates to a manufacturing control system comprising: a hardware processor; and a data collection unit configured to collect manufacturing data from a manufacturing process and communicate the manufacturing data to the hardware processor.
  • the manufacturing control system may be adapted to perform the presently disclosed manufacturing process for medicinal products or medical devices.
  • the hardware processor may thereby by configured to:
  • the manufacturing data may be operational data related to equipment used in the manufacturing process, or related to properties of the goods, such as medicinal products or medical devices, produced in the manufacturing process.
  • the collected manufacturing data may also be related to the packaging or labelling of the goods in the manufacturing process.
  • the requirement generation system comprises at least one hardware processor.
  • a requirement generation model comprises an artificial neural network trained to convert a first set of requirements to a second set of requirements relating to a use, a technical specification or a technical qualification and/or commissioning of the industrial manufacturing, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities.
  • the requirement generation model may comprise a recursive neural network trained to convert a first set of requirements to a second set of requirements relating to a use, a technical specification or a technical qualification and/or
  • a recursive neural network is a kind of deep neural network created by applying the same set of weights recursively over a structure, to produce a structured prediction over variable-size input structures, or a scalar prediction on it, by traversing a given structure in topological order.
  • Such a network may be used to produce or update requirements with the presently disclosed method and system for qualification and commissioning.
  • a recurrent neural network is a class of artificial neural network where connections between units form a directed cycle. This allows it to exhibit dynamic temporal behavior.
  • a RNN has turned out to particularly useful for requirement generation in relation to industrial manufacturing, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities used in said manufacturing, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices.
  • a first convolutional neural network could be used to convert the first set of requirements to a second set of requirements.
  • a second convolutional neural network model would be trained to convert the first set of requirements to the second set of requirements based on confirmed training data.
  • convolutional neural network are used, wherein the first and second convolutional neural networks are configured to perform the tasks of the corresponding first and second recurrent neural networks.
  • the first and second recurrent neural networks may thus be replaced by equivalent functions or neural networks.
  • the first and second sets of requirements may be any requirements in the sense that the generation system and method may be used to generate requirements not only from for example a given user requirement to a functional specification, but may move from any one type of requirements to any other type of requirements, which can be seen as a particular strength of the system.
  • One advantage of this approach is that if a parameter is updated in the system, the consequences for any other requirement can be given. More concretely this means that in an example, in which a V-model is used as model of the validation of the manufacturing of a medicinal product, the generation of requirement does not have to follow the logical sequence of such a V-model. With the presently disclosed approach it is possible to generate requirements for any state in the V-model based on any other state of the V-model. This makes the approach very flexible.
  • the user requirements may be given as a starting point and used as a first set of requirements.
  • a first set of requirements comprising user requirement may therefore comprise specifications for the industrial manufacturing, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities used in said manufacturing, labelling, assembling and/or packaging process of medicinal products and/or medical devices.
  • user requirements may comprise user requirement specifications for equipment, facilities and/or utilities. Typically they may be derived from requirements associated with a specific machine or equipment or use of such machine or equipment, and/or derived from regulatory requirements.
  • the presently disclosed method for qualifying and commissioning and requirement generation system being more useful than for example a system based on other automated systems is that companies often apply their own interpretation of user requirements and/or have specific reasons for deliberately interpreting requirements stricter or looser than necessary.
  • the presently disclosed system and method have turned out to be particularly efficient in handling such scenarios.
  • the user requirements may furthermore comprise information and/or technical parameters related to security of the manufacturing, labelling, processing, assembling and/or packaging process. Requirements for achieving safe processes may be provided by machine equipment providers, by authorities or by any party carrying out the process.
  • the user requirement specifications may comprise a requirement stating that manufactured products may not be damaged in the process and/or that manufactured products may not be heated above a predefined temperature in the process and/or that manufactured products may not be shaken more than a predefined limit in the process and/or that manufactured products may not be scratched in the process.
  • the user requirements may serve as first set of requirements but may also be generated backwards, wherein the user requirements are then seen as second set of requirements.
  • Example of functional requirement specification include requirements on how manufactured products are transported or gripped in the manufacturing, labelling, processing, assembling and/or packaging process.
  • Functional requirement specification may include any kind requirements related to the physical handling of the product.
  • a full manufacturing, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices typically comprises one more preparation- oriented part of the process towards an implementation, and one more validation- oriented part of the process.
  • the second set of requirements is selected from installation qualification, operational qualification and performance qualification. This means that in this embodiment any requirement related to a qualification can be generated from e.g. a user requirement or a functional
  • An installation qualification requirement may comprises for example requirements on layout qualification and/or mechanical qualification of the equipment or facilities used in said manufacturing, labelling, assembling and/or packaging process of medicinal products and/or medical devices.
  • An operational qualification requirement may comprise for example a verification of whether defect manufactured products are correctly identified and/or whether at least a predefined level of stability of the equipment is obtained for a given intensity of operation.
  • a performance qualification may comprise a verification of whether further components or processes related to the manufacturing, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities used in said manufacturing, labelling, processing, assembling and/or packaging process are functional and/or whether manufactured products and material can be automatically handled outside the equipment or facilities.
  • a V-model can be seen as a model of the main steps to be taken in conjunction with the corresponding deliverables within computerized system validation framework, or project life cycle development. It describes the activities to be performed and the results that have to be produced during product development. It can be seen as a logical sequence for specifying, executing and qualifying a project.
  • the presently disclosed computer-implemented requirement generation system may be applied to a V-model.
  • the plurality of sets of requirements may therefore be selected from a logical sequence of verification and/or validation steps of the industrial manufacturing, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities used in said manufacturing, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices.
  • the logical sequence of verification and/or validation steps may correspond to a v-model of an automated manufacturing process of pharmaceuticals.
  • the functionality of the first recurrent neural network model preferably involves encoding of an input sequence. Therefore, in one embodiment the first recurrent neural network may be trained to read and convert an input sequence of source text, such as a variable-length input sequence, to a fixed-length vector representation of the encoded intermediate vector of numbers or characters. This input sequence may be used as the first set of requirements.
  • the inventor has realized that by treating requirements of different stages of an industrial manufacturing, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices as sequential input to a recurrent neural network surprisingly accurate and fast results can be achieved.
  • the second recurrent neural network model can be trained to convert the vector representation into a variable-length output sequence corresponding to the second set of requirements.
  • the first and second recurrent neural network models may be trained to generate the second set of requirements from the first set of requirements according to a number of confirmed previous generations of
  • the second recurrent neural network model may run iteratively, wherein the model is updated after each iteration to predict the second requirement in the sequence after each requirement generation.
  • the first set of requirements may comprise a sequence of characters or words, and the first recurrent neural network model may consequently be trained to read and convert the characters or words sequentially to generate the intermediate set of numbers or characters.
  • the first recurrent neural network model may also updated iteratively to predict a new part of the encoded intermediate vector of numbers or characters.
  • the presently disclosed computer-implemented requirement generation system for generating a set of requirements for an industrial manufacturing, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities used in said manufacturing, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices may be further arranged to perform automation of a validation lifecycle of an industrial manufacturing, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices based on the generated set of
  • the complexity and required time in such automation makes the presently disclosed system suitable for such application.
  • the system can be used not only to automate a process by applying new ways of approaching the data, but can also be further arranged to validate industrial equipment and/or an industrial manufacturing, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices based on the generated set of
  • one advantage of the presently disclosed method and system is that in contrast to existing workflows, wherein the workflow may be predefined in the sense that when a user manually generates requirements he works in a predefined order, the present system offers possibilities to generate requirement based in one category or for one purpose based on any other category or purpose.
  • the system also has the advantage that if any requirement is missing the system may still be capable of completing the requirement generation without any loss since a sufficiently trained model may be able to compute the missing requirement by analyzing the context.
  • the system is further arranged to provide a replacement for an unknown requirement of the first set of requirements.
  • the presently disclosed system and method may therefore further comprise the step of training the relevant models while the process is carried out, or, alternatively, as a separate process.
  • Such training may involve for example the generation of a database of ground truth datasets, which can be considered to represent true and/or correct data.
  • a first and second recurrent neural network are used the training can be made for one of the models or both in a combined training.
  • the system is arranged to obtain a first training corpus comprising validated encoding data and wherein the first recurrent neural network model is trained to encode requirements of the first set of requirements to the encoded intermediate vector of numbers or characters based on the first training corpus.
  • system is arranged to obtain a second training corpus comprising validated decoding data and wherein the second recurrent neural network model is trained to decode the encoded intermediate vector of numbers or characters to the second set of requirements based on the second training corpus.
  • the computer-implemented requirement generation system comprises a plurality of sets of requirements relating to a use, a technical specification or a technical qualification and/or commissioning of the industrial manufacturing, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities; a first recurrent neural network model trained to encode requirements of a first set of requirements to an encoded intermediate vector of numbers or characters; and a second recurrent neural network model trained to decode the encoded intermediate vector of numbers or characters to a second set of requirements.
  • the first set of requirements may be seen from the system perspective as sequential data.
  • the first set of requirement may thereby be provided as a sequence of raw input, such as a comma separated file, wherein one requirement corresponds to one sequential input item.
  • the encoding and decoding to/from a vector of numbers and/or characters may thereby be carried out word by word on a lower level, and/or sentence by sentence on a higher level.
  • Fig. 1 shows an embodiment of the presently disclosed method (100) for qualifying and/or commissioning a process and/or equipment, wherein, in a first step (101 ), a first set of requirements related to a use, a technical specification or a technical qualification and/or commissioning of the industrial manufacturing, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities is obtained. In a second step (102) a first recurrent neural network model trained to encode requirements to a vector of numbers or characters is obtained. In a third step (103) the first set of requirements is encoded to a vector of numbers or characters.
  • a second recurrent neural network model trained to decode the encoded intermediate vector of numbers or characters to a second set of requirements is obtained.
  • the vector of numbers or characters is decoded to a second set of requirements.
  • a qualification and/or commissioning may be carried out.
  • Fig. 2 shows a schematic illustration of an embodiment of the presently disclosed computer-implemented requirement generation system (200) for generating a set of requirements for an industrial process and/or equipment.
  • the system (200) comprises a requirement generation model (202) having first and a second recurrent neural network (203, 204), a hardware processor (201 ), a storage device (205) for storing the instructions for carrying out the presently disclosed method, and may additionally comprise user interfaces, external resources (206) etc.
  • Fig. 3 shows an example of a setup, wherein two recurrent neural networks (203, 204) are arranged in a sequence-to-sequence configuration for operating according to the presently disclosed computer-implemented requirement generation system and qualification/commissioning method.
  • a first set of requirements (207) is encoded using the first RNN (203) into a vector of numbers or characters (208). Subsequently the vector of numbers or characters (208) is decoded into a second set of requirements (209).
  • Fig. 4 shows an example of groups of requirements according to a V-model (300).
  • the V-model shown in fig. 4 is not the only possible V-model for carrying out the invention. Any V-model or other representation of development lifecycle of a system may be used.
  • the example of V-model of fig. 4 comprises user requirement specifications (301 ), functional design specifications (302), detailed design specifications (303), implementation (304), as well as installation qualification (305), operational qualification (306) and performance qualification (307).
  • Fig. 5 illustrates that different types of requirements may influence each other and that requirements from different sources, in this case machine vendor requirements (401 ), regulatory requirements (402) and manufacturing operation procedure requirements (403), may imply a combined limited scope of final requirements.
  • Fig. 6 shows an embodiment of the presently disclosed manufacturing process for medicinal products or medical devices (500), wherein, in a first step (501 ), a first set of requirements is obtained. In a second step (502) the first set of requirements is sequentially encoded to a vector of numbers or characters. In a third step (503) the encoded intermediate vector of numbers or characters is decoded to a second set of requirements. In a fourth step (504) the step of manufacturing the medicinal products or medical devices is performed. In a fifth step (505) it is validated that the second set of requirements is fulfilled in the step of manufacturing the medicinal products or medical devices.
  • Fig. 7 shows a schematic illustration of an embodiment of the presently disclosed manufacturing control system (600).
  • the system (600) comprises a requirement generation model (604) having first and a second recurrent neural network (605, 606), a hardware processor (601 ), a storage device (603) for storing the instructions for carrying out the method, and a data collection unit (602).
  • a computer-implemented requirement generations system for generating a set of requirements for an industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities used in said manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, the requirement generation system comprising at least one hardware processor and at least one storage device storing requirement generation instructions and a requirement generation model,
  • requirement generation model comprises an artificial neural network trained to generate
  • a method for qualifying and/or commissioning an industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities related to said process comprising the steps of:
  • a computer-implemented requirement generation system for generating a set of requirements for an industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities used in said manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, the requirement generation system comprising at least one hardware processor and at least one storage device storing requirement generation instructions and a requirement generation model,
  • requirement generation model comprises:
  • the requirement generation instructions when executed by the at least one hardware processor, causes the requirement generation system to generate the second set of requirements based on an update of, or the provision of a new, first set of requirements.
  • the first set of requirements comprises user requirement specifications for the industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities used in said manufacturing, formulation and filling, labelling, assembling and/or packaging process of medicinal products and/or medical devices.
  • the computer-implemented requirement generation system according to any of the preceding items, wherein the first set of requirements comprises user requirement specifications for equipment, facilities and/or utilities.
  • the computer-implemented requirement generation system according to any of items 4-5, wherein the user requirement specifications are derived from requirements associated with a specific machine or equipment or use of such machine or equipment, and/or derived from regulatory requirements.
  • the computer-implemented requirement generation system according to any of items 4-6, wherein the user requirement specifications comprises information related to security of the manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process.
  • the user requirement specifications comprises a requirement that manufactured products may not be damaged in the process and/or that manufactured products may not be heated above a predefined temperature in the process and/or that manufactured products may not be shaken more than a predefined limit in the process and/or that manufactured products may not be scratched in the process.
  • the computer-implemented requirement generation system according to any of the preceding items, wherein the first set of requirements or the second set of requirements comprises functional requirement specifications for the
  • the computer-implemented requirement generation system according to item 9, wherein the functional requirement specifications comprise requirements on how manufactured products are transported or gripped in the manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process.
  • the computer-implemented requirement generation system according to any of the preceding items, wherein the second set of requirements is selected from installation qualification, operational qualification and performance qualification.
  • the computer-implemented requirement generation system according to item 1 1 , wherein the installation qualification comprises layout qualification and/or mechanical qualification of the equipment or facilities used in said manufacturing, formulation and filling, labelling, assembling and/or packaging process of medicinal products and/or medical devices.
  • the computer-implemented requirement generation system according to any of items 1 1 -12, wherein the operational qualification comprises a verification of whether defect manufactured products are correctly identified and/or whether at least a predefined level of stability of the equipment is obtained for a given intensity of operation.
  • the computer-implemented requirement generation system according to any of items 1 1 -13, wherein the performance qualification comprises a verification of whether further components or processes related to the manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities used in said manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process are functional and/or whether manufactured products and material can be automatically handled outside the equipment or facilities.
  • the computer-implemented requirement generation system according to any of the preceding items, wherein the plurality of sets of requirements are selected from a logical sequence of verification and/or validation steps of the industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities used in said manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices.
  • the computer-implemented requirement generation system according to item 15 wherein the logical sequence of verification and/or validation steps correspond to a v-model of an automated manufacturing process of
  • the computer-implemented requirement generation system according to any of the preceding items, wherein the first recurrent neural network model is trained to read and convert an input sequence of source text, such as a variable-length input sequence, to a fixed-length vector representation of the encoded intermediate vector of numbers or characters.
  • the computer-implemented requirement generation system according to any of the preceding items, wherein the first and second recurrent neural network models are trained to generate the second set of requirements from the first set of requirements according to a number of confirmed previous generations of corresponding requirements.
  • the second recurrent neural network model is trained to generate a sequence of requirements, such as a sequence of words, from the encoded intermediate vector of numbers or characters by predicting a next requirement one by one based on a state of the second recurrent neural network model.
  • the first set of requirements comprises a sequence of characters or words and wherein the first recurrent neural network model is trained to read and convert the characters or words sequentially to generate the intermediate set of characters.
  • the computer-implemented requirement generation system according to any of the preceding items arranged to perform automation of a validation lifecycle of an industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices based on the generated set of requirements.
  • the computer-implemented requirement generation system according to any of the preceding items further arranged to validate industrial equipment and/or an industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices based on the generated set of requirements.
  • the computer-implemented requirement generation system according to any of the preceding items, further arranged to provide a replacement for an unknown requirement of the first set of requirements.
  • the computer-implemented requirement generation system according to any of the preceding items, wherein said system is arranged to obtain a first training corpus comprising validated encoding data and wherein the first recurrent neural network model is trained to encode requirements of the first set of requirements to the encoded intermediate vector of numbers or characters based on the first training corpus.
  • the computer-implemented requirement generation system according to any of the preceding items, wherein said system is arranged to obtain a second training corpus comprising validated decoding data and wherein the second recurrent neural network model is trained to decode the encoded intermediate vector of numbers or characters to the second set of requirements based on the second training corpus.
  • the computer-implemented requirement generation system according to any of the preceding items, wherein the first set of requirement is provided as a sequence of raw input, such as a comma separated file, wherein one requirement corresponds to one sequential input item.
  • the computer-implemented requirement generation system according to any of the preceding items, wherein the encoding and decoding are carried out word by word.
  • the computer-implemented requirement generation system according to any of the preceding items, wherein the encoding and decoding are carried out sentence by sentence.
  • the computer-implemented requirement generation system according to any of the preceding items the system being configured to carry out the requirement generation instructions automatically in substantially real-time.
  • the computer-implemented requirement generation system according to any of the preceding items the system being further configured to control the industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities used in said manufacturing, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices.
  • a method for qualifying and/or commissioning an industrial manufacturing, formulation and filling, labelling, processing, assembling and/or packaging process of medicinal products and/or medical devices, and/or equipment or facilities related to said process said method comprising the steps of:

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Development Economics (AREA)
  • Finance (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Human Resources & Organizations (AREA)
  • Accounting & Taxation (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Medical Treatment And Welfare Office Work (AREA)
  • Medical Preparation Storing Or Oral Administration Devices (AREA)

Abstract

La présente invention concerne un procédé de fabrication de produits médicinaux ou de dispositifs médicaux, ledit procédé comprenant les étapes consistant à : obtenir un premier ensemble d'exigences liées à des premières caractéristiques techniques de ladite fabrication de produits médicinaux ou de dispositifs médicaux ; coder séquentiellement le premier ensemble d'exigences en un vecteur intermédiaire codé de nombres ou de caractères à l'aide d'un premier modèle de réseau neuronal récurrent ou équivalent ; décoder séquentiellement le vecteur intermédiaire codé de nombres ou de caractères à un second ensemble d'exigences liées à des secondes caractéristiques techniques de ladite fabrication de produits médicinaux ou de dispositifs médicaux, à l'aide d'un second modèle de réseau neuronal récurrent ou équivalent ; réaliser l'étape de fabrication des produits médicinaux ou des dispositifs médicaux sur la base d'au moins le second ensemble d'exigences ; et valider que le second ensemble d'exigences est satisfait dans l'étape de fabrication des produits médicinaux ou des dispositifs médicaux.
PCT/EP2018/080018 2017-11-03 2018-11-02 Procédé de fabrication, procédé de qualification et/ou de mise en service d'une fabrication industrielle, et qualification et mise en service d'équipement et de processus WO2019086610A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
EP17199810.7 2017-11-03
EP17199810 2017-11-03

Publications (1)

Publication Number Publication Date
WO2019086610A1 true WO2019086610A1 (fr) 2019-05-09

Family

ID=60327058

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2018/080018 WO2019086610A1 (fr) 2017-11-03 2018-11-02 Procédé de fabrication, procédé de qualification et/ou de mise en service d'une fabrication industrielle, et qualification et mise en service d'équipement et de processus

Country Status (1)

Country Link
WO (1) WO2019086610A1 (fr)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH0394365A (ja) * 1989-09-07 1991-04-19 Omron Corp ニューラルネットワークシステム
US5432887A (en) * 1993-03-16 1995-07-11 Singapore Computer Systems Neural network system and method for factory floor scheduling
US20160328644A1 (en) * 2015-05-08 2016-11-10 Qualcomm Incorporated Adaptive selection of artificial neural networks
US20170140753A1 (en) * 2015-11-12 2017-05-18 Google Inc. Generating target sequences from input sequences using partial conditioning

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH0394365A (ja) * 1989-09-07 1991-04-19 Omron Corp ニューラルネットワークシステム
US5432887A (en) * 1993-03-16 1995-07-11 Singapore Computer Systems Neural network system and method for factory floor scheduling
US20160328644A1 (en) * 2015-05-08 2016-11-10 Qualcomm Incorporated Adaptive selection of artificial neural networks
US20170140753A1 (en) * 2015-11-12 2017-05-18 Google Inc. Generating target sequences from input sequences using partial conditioning

Similar Documents

Publication Publication Date Title
Gurbuz et al. Model-based testing for software safety: a systematic mapping study
JP6621204B2 (ja) 安全重視ソフトウェア開発のためのモデルベース技術および過程のためのシステムおよび方法
CN109492402A (zh) 一种基于规则引擎的智能合约安全评测方法
Fernandez-Llatas et al. Using process mining for automatic support of clinical pathways design
Singh Using Event-B for Critical Device Software Systems
Gonzalez et al. Verifying GSM-based business artifacts
US20140324908A1 (en) Method and system for increasing accuracy and completeness of acquired data
CN109614103A (zh) 一种基于字符的代码补全方法及系统
Novák et al. Digitalized automation engineering of Industry 4.0 production systems and their tight cooperation with digital twins
Gradišnik et al. Impact of historical software metric changes in predicting future maintainability trends in open-source software development
Bagheri et al. Synthesis of assurance cases for software certification
Wilking et al. Integrating machine learning in digital twins by utilizing sysml system models
Malins et al. SysML activity models for applying ISO 14971 medical device risk and safety management across the system lifecycle
Sydorov et al. Development of an approach to using a style in software engineering
Franceschetti et al. ProAmbitIon: Online Process Conformance Checking with Ambiguities Driven by the Internet of Things.
Singh et al. Formal verification and code generation for solidity smart contracts
WO2019086610A1 (fr) Procédé de fabrication, procédé de qualification et/ou de mise en service d'une fabrication industrielle, et qualification et mise en service d'équipement et de processus
da Silva Oliveira et al. Obtaining formal models from ladder diagrams
Singh et al. Use of tabular expressions for refinement automation
Singh et al. A formal approach to rigorous development of critical systems
EP3633561A1 (fr) Système et procédé pour gérer des séquences d'éléments de contrainte
JP6290147B2 (ja) 制御機器プログラムコードを作成するコンピュータ実装方法および関連するメッセージ管理システム
Bowen et al. Formality, Agility, Security, and Evolution in Software Development.
Roussel et al. Design of logic controllers thanks to symbolic computation of simultaneously asserted Boolean equations
Oueslati et al. Distributed reconfigurable b approach for the specification and verification of b-based distributed reconfigurable control systems

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18793438

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18793438

Country of ref document: EP

Kind code of ref document: A1