Μοντελοποίηση θεμελιώδους σχεδιασμού βασισμένη στην τεχνητή νοημοσύνη - Εφαρμογή στο μηχανοτρονικό σχεδιασμό (Μουλιανίτης, 2003)

Περίληψη:

Ο σχεδιασμός, ως ανθρώπινη ενασχόληση για πολλούς αιώνες, στηρίχθηκε κυρίως στην εμπειρία και στο ταλέντο. Μετά τα μέσα του 20ου αιώνα, αναπτύχθηκαν επιστημονικές προσπάθειες ώστε ο σχεδιασμός να αποτελέσει αυτόνομο επιστημονικό πεδίο. Από τότε αναπτύχθηκαν αρκετές φιλοσοφίες, μοντέλα, μέθοδοι και τεχνικές για το σχεδιασμό. Επειδή ο σχεδιασμός είναι μια νοητική εργασία με ειδικά χαρακτηριστικά για κάθε άτομο ή ομάδα, η συστηματική του οργάνωση είναι πολύ δύσκολη. Η θεωρητική θεμελίωση του σχεδιασμού είναι ακόμη δυσκολότερη και μέχρι τώρα έχουν εμφανιστεί λίγες απόπειρες για την ανάπτυξη μιας γενικευμένης θεωρίας συστηματικού σχεδιασμού, η οποία βρίσκεται σε προεπιστημονικό στάδιο, ενώ έχει αναπτυχθεί πλούσια γνώση που αναφέρεται σε μεθοδολογίες σχεδιασμού καθώς και εφαρμογές. Στη διατριβή αυτή εισάγεται ένα μαθηματικό μοντέλο που συνεισφέρει στη συστηματικοποίηση του θεμελιώδους σχεδιασμού. Ερμηνεύονται με μαθηματικό τρόπο οι γενικές διαδικασίες παραγωγής και αξιολόγησης των λύσεων στο θεμελιώδη σχεδιασμό, και μοντελοποιούνται τα αντικείμενα και οι κανόνες χρησιμοποιώντας ασαφή ή δίτιμη ή πολύτιμη λογική ανάλογα με το πόσο καλά καθορισμένο είναι το πρόβλημα που πρέπει να επιλυθεί. Στη διατριβή αναπτύχθηκε ένα μαθηματικό μοντέλο για τη διαδικασία παραγωγής λύσεων στο θεμελιώδη σχεδιασμό. Αναπτύχθηκε ένας μηχανισμός για την αυτόματη παραγωγή βαθμών ικανοποίησης των εφικτών λύσεων ως προς τις σχεδιαστικές προδιαγραφές και των σχεδιαστικών περιορισμών. Το σύνολο των υποψηφίων τελικών λύσεων παράγεται από ένα πίνακα που περιέχει τους βαθμούς αυτούς. Η παραγωγή του συνόλου των τελικών λύσεων που είναι ένα υποσύνολο του συνόλου των υποψηφίων τελικών λύσεων πραγματοποιείται χρησιμοποιώντας το Ευκλείδειο μέτρο. Επειδή, λόγω συνδυαστικής έκρηξης, η αναζήτηση των λύσεων μερικές φορές είναι χρονοβόρα, αναπτύχθηκαν δύο ευρετικές μέθοδοι οι οποίες στηρίζονται σε στοιχεία διανυσματικής ανάλυσης για να επιταχυνθεί αυτή η διαδικασία. Εισάγεται επίσης ένα γενικευμένο πλαίσιο κατασκευής δεικτών για την αξιολόγηση των θεμελιωδών λύσεων. Για την κατασκευή των δεικτών χρησιμοποιούνται τελεστές της ασαφούς λογικής και πιο συγκεκριμένα οι T-norms και οι τελεστές μέσου όρου. Αναπτύχθηκαν επίσης οδηγίες για την επιλογή της κλάσης των τελεστών αλλά και για τους τελεστές αυτούς καθαυτούς. Ο σχεδιαστής μπορεί να κατασκευάσει δείκτες αξιολόγησης με τους οποίους μπορεί να αξιολογήσει και να επιλέξει θεμελιώδεις λύσεις με βάση ειδικά χαρακτηριστικά, όπως είναι η μηχανοτρονικότητα κ. άλ. που θέλει να κατέχει το τελικό προϊόν. Επίσης, εισήχθηκε ο “Μηχανοτρονικός Δείκτης” με τον οποίο ο σχεδιαστής μπορεί να επιλέξει τις πιο ευέλικτες, ευφυείς και με τη μικρότερη πολυπλοκότητα λύσεις. Η διαμόρφωση του δείκτη στηρίχθηκε στα τρία αυτά χαρακτηριστικά που κατέχουν τα μηχανοτρονικά προϊόντα. Αναπτύχθηκε ένα σύστημα βασισμένο σ’ αυτή τη γνώση για το μηχανοτρονικό σχεδιασμό ρομποτικών αρπάγων για το χειρισμό εύκαμπτων υλικών. Η απαραίτητη για τον σκοπό αυτό γνώση προήλθε από την σχετική βιβλιογραφία και από τους ειδικούς (experts) και κατόπιν οργανώθηκε κατάλληλα. Το σύστημα αυτό αποτελεί το πρώτο που εμφανίζεται στη βιβλιογραφία για το χειρισμό εύκαμπτων υλικών, και το πρώτο που υλοποιεί το θεμελιώδη σχεδιασμό αρπαγών. Στο σύστημα αυτό δοκιμάστηκαν και μελετήθηκαν οι παραπάνω μέθοδοι παραγωγής και αξιολόγησης λύσεων. Οι μέθοδοι που αναπτύσσονται στην παρούσα διατριβή μπορούν να χρησιμοποιηθούν για την πραγματοποίηση του μηχανοτρονικού θεμελιώδους σχεδιασμού ή για την ανάπτυξη νέων πιο αποτελεσματικών συστημάτων βασισμένων σε γνώση για το μηχανοτρονικό θεμελιώδη σχεδιασμό.

Link: http://openarchives.ekt.gr/visit/87867

 

Ιt was only by the middle of the 20th century when design – an intense human activity – became the focus point of scientists towards the direction of its formalization and “scientification”. Since then, design philosophies, models, methods and techniques have been developed. The systematic formulation of design is very difficult since it is a perceptive process with special characteristics for every person or team. The theoretical foundation of design is more difficult and until now there have been made relatively few attempts for developing a generalized theory of systematic design. This theory is still in pre-scientific phase, while there is rich knowledge on design methodologies in specific applications. In this thesis, a mathematical model towards systemizing the conceptual design phase is introduced. A mathematical interpretation of the general concept generation and evaluation processes is analyzed. The design entities and rules governing a design problem are represented using fuzzy, Boolean and many-valued logic depending on how well-defined is the knowledge of the field that the model is applied to. A mechanism for the automatic generation of the degrees of satisfaction of the design specifications by the feasible solutions subjected to the design constraints is also proposed. From a matrix that contains these degrees, the set of all possible combined solutions is generated. The generation of the set of final solutions, that is a subset of the set of the combined solutions, is based on a suitable partition of the Euclidean space. In order to avoid the combinatorial explosion that occurs during design space search, two heuristic search methods have been developed based on vector analysis. A frame for introducing design indices for the evaluation of alternative solutions is also proposed. The formulation of the design indices is based on fuzzy logic and more specific on T-norms and averaging operators. The guidelines for the selection of the class of operators and then for the operator itself are also given. Using this frame, the designer can form evaluation indices based on special characteristics that the product must inherit. Furthermore, the “Mechatronics Index” is introduced. Using this index, the designer can select the most intelligent and flexible alternative solutions that have the minimum complexity. The formulation of the index is based on these three common characteristics of the mechatronics products. Finally, the knowledge for the design of grippers for handling non-rigid materials is acquired from experts and literature and organized in a knowledge-based system. This system is the first to appear in the international literature for the conceptual design of grippers for handling non-rigid materials. In this system the above methods for the concept generation and evaluation has been tested and studied. The present thesis proposes a methodology that of it can be used for the realization of the conceptual mechatronics design or for the development of new intelligent knowledge-based systems for the conceptual mechatronics design of products. - See more at: http://nemertes.lis.upatras.gr/jspui/handle/10889/278#sthash.aMoHJcPI.dpuf

 

 

Ιt was only by the middle of the 20th century when design – an intense human activity – became the focus point of scientists towards the direction of its formalization and “scientification”. Since then, design philosophies, models, methods and techniques have been developed. The systematic formulation of design is very difficult since it is a perceptive process with special characteristics for every person or team. The theoretical foundation of design is more difficult and until now there have been made relatively few attempts for developing a generalized theory of systematic design. This theory is still in pre-scientific phase, while there is rich knowledge on design methodologies in specific applications. In this thesis, a mathematical model towards systemizing the conceptual design phase is introduced. A mathematical interpretation of the general concept generation and evaluation processes is analyzed. The design entities and rules governing a design problem are represented using fuzzy, Boolean and many-valued logic depending on how well-defined is the knowledge of the field that the model is applied to. A mechanism for the automatic generation of the degrees of satisfaction of the design specifications by the feasible solutions subjected to the design constraints is also proposed. From a matrix that contains these degrees, the set of all possible combined solutions is generated. The generation of the set of final solutions, that is a subset of the set of the combined solutions, is based on a suitable partition of the Euclidean space. In order to avoid the combinatorial explosion that occurs during design space search, two heuristic search methods have been developed based on vector analysis. A frame for introducing design indices for the evaluation of alternative solutions is also proposed. The formulation of the design indices is based on fuzzy logic and more specific on T-norms and averaging operators. The guidelines for the selection of the class of operators and then for the operator itself are also given. Using this frame, the designer can form evaluation indices based on special characteristics that the product must inherit. Furthermore, the “Mechatronics Index” is introduced. Using this index, the designer can select the most intelligent and flexible alternative solutions that have the minimum complexity. The formulation of the index is based on these three common characteristics of the mechatronics products. Finally, the knowledge for the design of grippers for handling non-rigid materials is acquired from experts and literature and organized in a knowledge-based system. This system is the first to appear in the international literature for the conceptual design of grippers for handling non-rigid materials. In this system the above methods for the concept generation and evaluation has been tested and studied. The present thesis proposes a methodology that of it can be used for the realization of the conceptual mechatronics design or for the development of new intelligent knowledge-based systems for the conceptual mechatronics design of products. - See more at: http://nemertes.lis.upatras.gr/jspui/handle/10889/278#sthash.aMoHJcPI.dpuf

 

 

Ιt was only by the middle of the 20th century when design – an intense human activity – became the focus point of scientists towards the direction of its formalization and “scientification”. Since then, design philosophies, models, methods and techniques have been developed. The systematic formulation of design is very difficult since it is a perceptive process with special characteristics for every person or team. The theoretical foundation of design is more difficult and until now there have been made relatively few attempts for developing a generalized theory of systematic design. This theory is still in pre-scientific phase, while there is rich knowledge on design methodologies in specific applications. In this thesis, a mathematical model towards systemizing the conceptual design phase is introduced. A mathematical interpretation of the general concept generation and evaluation processes is analyzed. The design entities and rules governing a design problem are represented using fuzzy, Boolean and many-valued logic depending on how well-defined is the knowledge of the field that the model is applied to. A mechanism for the automatic generation of the degrees of satisfaction of the design specifications by the feasible solutions subjected to the design constraints is also proposed. From a matrix that contains these degrees, the set of all possible combined solutions is generated. The generation of the set of final solutions, that is a subset of the set of the combined solutions, is based on a suitable partition of the Euclidean space. In order to avoid the combinatorial explosion that occurs during design space search, two heuristic search methods have been developed based on vector analysis. A frame for introducing design indices for the evaluation of alternative solutions is also proposed. The formulation of the design indices is based on fuzzy logic and more specific on T-norms and averaging operators. The guidelines for the selection of the class of operators and then for the operator itself are also given. Using this frame, the designer can form evaluation indices based on special characteristics that the product must inherit. Furthermore, the “Mechatronics Index” is introduced. Using this index, the designer can select the most intelligent and flexible alternative solutions that have the minimum complexity. The formulation of the index is based on these three common characteristics of the mechatronics products. Finally, the knowledge for the design of grippers for handling non-rigid materials is acquired from experts and literature and organized in a knowledge-based system. This system is the first to appear in the international literature for the conceptual design of grippers for handling non-rigid materials. In this system the above methods for the concept generation and evaluation has been tested and studied. The present thesis proposes a methodology that of it can be used for the realization of the conceptual mechatronics design or for the development of new intelligent knowledge-based systems for the conceptual mechatronics design of products. - See more at: http://nemertes.lis.upatras.gr/jspui/handle/10889/278#sthash.aMoHJcPI.dpuf