This thesis is proposed as a practical answer to a frequent business need to have tool for optimizing the performance of the production system, such as the improvement of the total production time and the saturation of resources. The present paper starts with an analysis of the production system of a concrete business case, conducted with the help of management and specialized technical company personnel, in order to create a software for production scheduling. First, it was necessary to outline the mapping of the production system with the aim of identifying all production operations and their sequence. Secondly, the production constraints that characterize this system have been identified and analyzed. During the analysis, the contribution of the technical and operating personnel was fundamental, as was the study of the technical documents present in the company. The definition of the production system with the precise identification of the constraints that characterize it has allowed us to fully highlight the complexity of the system. In this context it was considered that the definition of an appropriate scheduling software represented a key choice for optimizing the production system. The analysis of the academic literature made it possible to identify some similar cases by construction and constraints in the case examined here. Literature in fact proposes different types of models for scheduling job shop type production systems: the classics, heuristics and meta-heuristics. The first, although advantageous for the immediate usability of the proposed solution, do not lend themselves to being applied in complex production contexts, considering the simplifying hypotheses that characterize them that are difficult to find in concrete cases. Even the latter prove to be less responsive to the needs of the business case in question, as even though they provide an optimized solution on the field, they do not allow scheduling forecasting on the entire production system. Meta-heuristic models, on the other hand, are well applicable to complex scheduling cases, as they allow, through sub-optimal solutions, to obtain an overall forecasting production scheduling while not being subject to simplistic hypotheses. Therefore, this work, considering the characteristics of the examined case, is based on the logics of meta-heuristic models, in particular preferring the genetic algorithm, coherently with the scientific literature that in recent years proposes it as one of the most performing methods for scheduling in productive systems . The software structure consists of a genetic algorithm and a simulation model. The first allows you to create a population whose individuals are sequences of schedulable jobs. The second reproduces the production system in a virtual way and processes the job sequences, organizing them in a time schedule. In this phase the interaction between the simulation model and the genetic algorithm allows to obtain a productive scheduling according to a fittest individuals optimization logic. In conclusion, the application of the software has demonstrated its concrete utility with satisfactory results, as the solutions obtained satisfy all the present production constraints and frequently have proved to be better than those applied in reality. This work also lays the foundations for a further study on the effects that a modification of the parameters of the genetic model would entail as well as on the possibility of a significant reduction in the time required to reach the optimal scheduling solution.
Questa tesi si propone come una risposta pratica ad una frequente esigenza aziendale di disporre di un tool informatico per l’ottimizzazione delle performance del sistema produttivo, come ad esempio il miglioramento del tempo totale di produzione e la saturazione delle risorse. Il presente elaborato prende avvio da un’analisi del sistema produttivo di un caso aziendale concreto, condotta con l’ausilio del management e del personale tecnico specializzato aziendale, al fine di realizzare un software per la schedulazione della produzione. In primis, è stato necessario delineare la mappatura del sistema produttivo con l’obiettivo di individuare tutte le operazioni produttive e la loro sequenza. In secundis, si è proceduto ad individuare ed analizzare i vincoli produttivi che caratterizzano tale sistema. Durante l’analisi l’apporto del personale tecnico operativo si è reso fondamentale, così come lo studio dei documenti tecnici presenti in azienda. La definizione del sistema produttivo con l’individuazione precisa dei vincoli che lo caratterizzano ha permesso di evidenziare in modo completo la complessità del sistema. In questo contesto si è ritenuto che la definizione di un appropriato software di scheduling rappresentasse una scelta chiave per ottimizzare il sistema produttivo. L’analisi della letteratura accademica ha consentito di individuare alcuni casi similari per costruzione e vincoli al caso qui preso in esame. La letteratura infatti propone diverse tipologie di modelli per la schedulazione di sistemi produttivi di tipo job shop: i classici, gli euristici e i meta-euristici. I primi seppur vantaggiosi per l’immediata fruibilità della soluzione proposta mal si prestano ad essere applicati in contesti produttivi complessi, considerando le ipotesi semplificative che li caratterizzano difficilmente riscontrabili nei casi concreti. Anche i secondi si rivelano poco rispondenti alle esigenze del caso aziendale in esame, in quanto pur fornendo una soluzione ottimizzata on the field, non consentono una previsione di scheduling sull’intero sistema di produzione. I modelli meta euristici invece risultano ben applicabili a casi di schedulazione complessa, in quanto consentono attraverso soluzioni sub-ottimali di ottenere uno scheduling produttivo previsionale complessivo pur non soggiacendo ad ipotesi semplicistiche. Pertanto, questo lavoro considerate le caratteristiche del caso esaminato si fonda sulle logiche dei modelli meta euristici, prediligendo in particolare l’algoritmo genetico, coerentemente con la letteratura scientifica che negli ultimi anni lo propone come uno dei metodi più performanti per la schedulazione in sistemi produttivi. La struttura del software si compone di un algoritmo genetico e di un modello di simulazione. Il primo consente di creare una popolazione i cui individui sono delle sequenze di job schedulabili. Il secondo riproduce in maniera virtuale il sistema produttivo e processa le sequenze di job organizzandole in una pianificazione temporale. In questa fase l’interazione tra il modello di simulazione e l’algoritmo genetico consente di ottenere uno scheduling produttivo secondo una logica di ottimizzazione fittest individuals. In conclusione, l’applicazione del software ha dimostrato la sua concreta utilità con risultati soddisfacenti, in quanto le soluzioni ottenute soddisfano tutti i vincoli produttivi presenti e frequentemente si sono dimostrate migliori rispetto a quelle applicate nella realtà. Questo lavoro pone altresì le basi per un ulteriore approfondimento sugli effetti che una modifica dei parametri del modello genetico comporterebbe nonché sulla possibilità di sensibile riduzione dei tempi per il raggiungimento della soluzione di scheduling ottimale.
Production scheduling optimization in a complex industrial case
CAMPAGNOLO, FILIPPO
2018/2019
Abstract
This thesis is proposed as a practical answer to a frequent business need to have tool for optimizing the performance of the production system, such as the improvement of the total production time and the saturation of resources. The present paper starts with an analysis of the production system of a concrete business case, conducted with the help of management and specialized technical company personnel, in order to create a software for production scheduling. First, it was necessary to outline the mapping of the production system with the aim of identifying all production operations and their sequence. Secondly, the production constraints that characterize this system have been identified and analyzed. During the analysis, the contribution of the technical and operating personnel was fundamental, as was the study of the technical documents present in the company. The definition of the production system with the precise identification of the constraints that characterize it has allowed us to fully highlight the complexity of the system. In this context it was considered that the definition of an appropriate scheduling software represented a key choice for optimizing the production system. The analysis of the academic literature made it possible to identify some similar cases by construction and constraints in the case examined here. Literature in fact proposes different types of models for scheduling job shop type production systems: the classics, heuristics and meta-heuristics. The first, although advantageous for the immediate usability of the proposed solution, do not lend themselves to being applied in complex production contexts, considering the simplifying hypotheses that characterize them that are difficult to find in concrete cases. Even the latter prove to be less responsive to the needs of the business case in question, as even though they provide an optimized solution on the field, they do not allow scheduling forecasting on the entire production system. Meta-heuristic models, on the other hand, are well applicable to complex scheduling cases, as they allow, through sub-optimal solutions, to obtain an overall forecasting production scheduling while not being subject to simplistic hypotheses. Therefore, this work, considering the characteristics of the examined case, is based on the logics of meta-heuristic models, in particular preferring the genetic algorithm, coherently with the scientific literature that in recent years proposes it as one of the most performing methods for scheduling in productive systems . The software structure consists of a genetic algorithm and a simulation model. The first allows you to create a population whose individuals are sequences of schedulable jobs. The second reproduces the production system in a virtual way and processes the job sequences, organizing them in a time schedule. In this phase the interaction between the simulation model and the genetic algorithm allows to obtain a productive scheduling according to a fittest individuals optimization logic. In conclusion, the application of the software has demonstrated its concrete utility with satisfactory results, as the solutions obtained satisfy all the present production constraints and frequently have proved to be better than those applied in reality. This work also lays the foundations for a further study on the effects that a modification of the parameters of the genetic model would entail as well as on the possibility of a significant reduction in the time required to reach the optimal scheduling solution.File | Dimensione | Formato | |
---|---|---|---|
Tesi Filippo Campagnolo.pdf
non accessibile
Descrizione: Testo della tesi
Dimensione
4.22 MB
Formato
Adobe PDF
|
4.22 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/148794