A falta de informação e de confiança entre os agentes em sistemas de recomendação que lidam com domínios dinâmicos podem ser fatores que contribuem para que os agentes gerem resultados de baixa qualidade. Na falta de informação para gerar recomendações, é necessário que os agentes sejam capazes de assumir ou compartilhar informações, criem laços de confiança entre si e que se adaptem às mudanças do estado do conhecimento para que sejam capazes de resolver os problemas. Esta tese apresenta a abordagem MATRES - uma abordagem multiagente baseada em suposições com mecanismo de confiança aplicada em um sistema de recomendação multiagente. Na abordagem MATRES, os agentes são capazes de lidar com conhecimento distribuído. Cada agente trabalha como especialista e é capaz de compartilhar seu conhecimento com os demais, de acordo com seus índices de confiança. Para a solução de um problema, diferentes tarefas são distribuídas entre os agentes. Algumas tarefas apresentam uma relação de dependência, fazendo com que uma tarefa dependa do resultado de outra. Nesta situação, o agente possui um componente de manutenção da verdade que permite a utilização de suposições para a realização das tarefas de forma assíncrona. Na falta de informação proveniente de outra tarefa, o agente é capaz de manipular suposições, sendo capaz de executar sua tarefa. Além disto, o componente de manutenção da verdade auxilia na manutenção da integridade das bases de conhecimento dos agentes. A abordagem MATRES foi validada em um cenário de recomendação de pacotes turísticos. Casos reais de uma agência de viagem foram utilizados na validação da abordagem e os resultados obtidos corroboram a hipótese de que que a abordagem proposta aumenta a assertividade das recomendações geradas pelos agentes em ambientes distribuídos e dinâmicos. / The lack of trust and information among agents in dynamic domains may contribute to the generation of poor results in multiagent recommender systems. These domains requires that agents exchange information, establishing bonds of trust among themselves and adapting the modification of the status of the knowledge to be able to solve problems. In systems where the knowledge is distributed among several agents, the exchange of information is essential for improving the performance of the agents and maybe leading to inconsistencies when the information exchanged has different status. This thesis presents the MATRES approach - a multiagent Assumption-Based recommender approach with a trust mechanism. In this approach agents are able to deal with distributed knowledge. Each agent works as an expert and is able to share its knowledge with other agents, according to its trust degree. In order to solve a problem, different tasks are distributed among the agents. Some tasks are interdependent, which means that to solve a task it is necessary to use the result from other one. In this situation, the agent has a truth maintenance component that allows using assumptions to perform tasks in a assynchronous ways and helps the maintenance of the integrity of the knowledge bases of the agents. TheMATRES approach was validated in the travel recommendation scenario. The results show that the proposal increases the assertiveness of the recommendations provided by the agents in this dynamic domain.

In today’s manufacturing industry, addressing maintenance functions is essential to increase effectiveness and productivity. With the present era of digitalisation and technology, all data is enabled to be shared and integrated within the organisations. Increased advancement of complex productions requires consequently greater performance maintenance systems leading to challenges in monitoring processes. Recent research indicates that there are deficiencies regarding maintenance when connecting technical- and organisational aspects. Therefore, it becomes engaging to investigate how to support the decision-making and how to achieve predictability by anticipating unplanned disruptions, breakdowns and failures. The study was commenced to analyse how organisations can utilise maintenance information in order to gain predictability. The research was conducted by using a qualitative approach gathering data through unobtrusive measures, interviews and observations, all being performed at a production manufacturing site. The collected data allowed the research questions to be answered, showing which maintenance information becomes critical to enable prediction in order to prevent or eliminate equipment breakdowns as well as presenting how the information should be utilised within the organisation.  The findings emphasise that in order to achieve predictability and preventive measures, organisations must assess the specific dimensions of downtime, severity and frequency. Both planned- and unplanned maintenance information should be logged with rich descriptions, having common inputs in all systems to induce data transferability. It is justified that systems should be integrated and not individually used, inducing the absolute necessity of standards. All involved individuals with distinct roles and responsibilities should have clear apprehensions about the necessity of maintenance as well as how it brings value. This supports the view of motivation where recognition and achievement become crucial to increase performance efforts, encouraging correct and detailed system inputs. Additionally, mutual understandings that all agree upon lead to maintenance systems becoming an influential organisational function that fully supports the decision-making. / I dagens tillverkningsindustri är det viktigt att ta hänsyn till underhållsfunktioner för att öka effektiviteten och produktiviteten. Med den nuvarande eran av digitalisering och teknik tillåts all data att delas och integreras i samtliga nivåer i organisationer. Med ökad utveckling av komplexa produktionssystem krävs det medföljande högre presterande underhållssystem som leder till utmaningar i processövervakning. Ny forskning tyder på att det finns brister med avseende på underhåll för att integrera tillsammans tekniska och organisationella aspekter. Av den anledning blir det engagerande att undersöka hur beslutsfattandet kan stödjas för att uppnå förutsägbarhet genom att förutse oplanerade störningar, haveri och fel. Studien inleddes för att analysera hur organisationer kan använda underhållsinformation för att erhålla förutsägbarhet. Utförandets design implementerades med hjälp av ett kvalitativt tillvägagångssätt genom att samla in data via diskret-forskning, intervjuer och observationer, allt utfört på en produktionsanläggning. Samtliga insamlade uppgifter fullgjorde det möjligt att besvara forskningsfrågorna om vilken underhållsinformation som blir avgörande för att förutsäga, förebygga eller eliminera utrustningsstörningar och hur tillhörande information ska nyttjas inom organisationen. Som resultatet betonar, för att erhålla förutsägbarhet och förebyggande åtgärder krävs det beaktning av dimensionerna stilleståndstid, svårighetsgrad och frekvens. Både planerat- och oplanerat underhållsarbete ska loggas med omfattande beskrivningar och gemensamma koder i alla system för att inducera överförbarheten. Det motiveras att system ska integreras och inte användas enskilt, vilket föranleder nödvändigheten av standarder. Samtliga involverade individer med distinkta roller och ansvarsområden ska ha en tydlig förståelse över nödvändigheten av underhållssystem och hur den skapar värde. Detta stödjer synen på motivation, där igenkännande och vidareutveckling blir avgörande för att öka prestationsinsatserna såväl som uppmuntra till korrekta och tydliga systemkoder. Detta förklarar att ömsesidig förståelse som samtliga är överens om leder till att underhållssystem blir en inflytelserik organisationsfunktion som stödjer beslutsunderlag.

This diploma thesis concerns issues of predictive and condition based maintenance system of machine tools with using a vibrodiagnostics. It studies and researches its impacts through the basic processes of the maintenance system and characterizes the vibration diagnosis as its tool and mean. There is also described a process of putting condition based maintenance into practice in the practical part of the thesis. The development is realized at Siemens Ltd. Brno with all its requirements and aspects such as a maintenance audit which means the decision on the suitability of condition based maintenance within the current maintenance system, technical analysis as a part of introduction of vibration diagnosis and the practical example of acquiring, recording and assessment of measured vibration. Prior to the end the economic evaluation of the planned predictive maintenance system and the design of the general model of development and implementation of the maintenance system into practice are included.

This work is intended to investigate the Aircraft Technical Diagnostics Systems by focusing on Central Maintenance Systems and the Fault Detection and Isolation (FDI) process among the avionic components. A review about the Computer Control Systems and background about the Avionic Architecture is presented prior to introducing to most popular FDI method; model-based diagnosis. The discussions about the onboard FDI practices are followed by a maintenance hangar FDI perspective which was concluded as a result of the field research. The outcomes of the field research and pointing the “real” fault isolation are analyzed to point the practical needs of a hangar FDI tools. Subsequently, a proposal technique, which can improve fault isolation by preventing No-Fault-Found (NFF), is given by discussing the methods to implement it. Finally, the results of investigation and conclusions of analysis are presented. / Baigiamajame darbe lietuviškos anotacijos nepateikta.