A carregar...
Publicação
Controlo Estatístico do Processo: Estudo de caso numa Indústria Cervejeira
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 13.66 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
A qualidade, ao longo dos anos, tem vindo a ganhar destaque em muitas organizações, pelo
facto de existir uma maior exigência por parte dos clientes que desejam bons produtos a preços
razoavelmente baixos. É um investimento que gera um retorno importante, não através do
lucro diretamente, mas pela prevenção dos erros por exemplo, que a longo prazo se traduz
num crescimento sustentável para as empresas.
O presente trabalho tem como principal objetivo impulsionar a melhoria contínua do processo
produtivo através da prática do controlo estatístico do processo. A monitorização da quali-
dade dos sistemas de produção é fundamental, no entanto, a realidade diverge frequente-
mente da teoria e os processos industriais revelam-se muito mais complexos do que a sua
idealização. Por isso, os pressupostos para aplicar técnicas estatísticas, por exemplo as cartas
de controlo tradicionais, por vezes são desrespeitados, o que origina a um aumento do número
de falsos alarmes que consecutivamente conduz a conclusões erradas a cerca do processo. Nes-
tas situações, para entender corretamente o comportamento do processo, é importante estudá-
lo a fundo, através de diferentes ferramentas estatísticas, para que haja um tratamento ade-
quado dos dados.
A metodologia proposta foi implementada numa indústria de bebidas, a ECM - Empresa de
Cervejas da Madeira, a fim de conhecer e validar a sua aplicabilidade. O estudo incidiu sobre
a cerveja Coral Branca, em 5 caraterísticas: Álcool, Extrato Primitivo, Estabilidade de Espuma,
Turvação e Amargor. Esta metodologia decompõe-se em duas fases, em que na Fase I pre-
tende-se verificar se o processo se encontra sob controlo estatístico e se têm capacidade de
produzir conforme os requisitos estabelecidos, e posteriormente, na Fase II onde é feito um
acompanhamento constante do processo ao longo do tempo. Para isso, aplicaram-se cartas de
controlo univariadas e multivariadas, e a partir dos resultados foi efetuada uma análise com-
parativa entre o desemprenho de ambas as cartas.
A aplicação das diferentes cartas de controlo na Fase I permitiu evidenciar que o processo
apresenta uma variação para além do considerado normal ou expectável. Ainda, verificou-se
que nenhuma das caraterísticas apresenta capacidade de produzir de acordo com as especificações pré-definidas. Estes resultados são normais visto que este cenário é verificado
em muitos processos na vida real. No entanto, para melhorar o desempenho e evitar situações
indesejadas sugere-se a implementação de ações corretivas. Por outro lado, os resultados al-
cançados na Fase II não foram os esperados pois foram comprometidos pela relação das séries
de dados de ambas as fases. Posto isto, não foi possível retirar conclusões especificas a cerca
do comportamento do processo nesta fase, nem comparar a performance entre o estudo uni-
variado e multivariado como era pretendido.
Over the years, quality has been gaining prominence in many organizations, since there are greater demands from customers who want good products at reasonably low prices. It's an investment that generates an important return, not through profit directly, but through the prevention of errors, for example, which in the long term translates into sustainable growth for companies. The main aim of this work is to drive continuous improvement in the production process through the practice of statistical process control. Monitoring the quality of production sys- tems is fundamental, but reality often diverges from theory and industrial processes turn out to be much more complex than idealized. For this reason, the assumptions for applying statis- tical techniques, such as traditional control charts, are sometimes disrespected, which leads to an increase in the number of false alarms and, consequently, to erroneous conclusions about the process. In these situations, in order to correctly understand the behavior of the process, it is important to study the process in depth, using different statistical tools, so that the data can be properly processed. The proposed methodology was implemented in a beverage industry, ECM - Empresa de Cer- vejas da Madeira, to understand and validate its applicability. The study focused on Coral Branca beer, in 5 characteristics: Alcohol, Original Extract, Foam Stability, Cloudiness and Bitterness. This methodology is divided into two phases: Phase I aims to check whether the process is under statistical control and if has the capacity to produce according to the established re- quirements, and then Phase II, where the process is constantly monitored over time. To this end, univariate and multivariate control charts were applied so that the results could be used to make a comparative analysis of the performance of both charts. The application of the different control charts in Phase I showed that the process varies beyond what is considered normal or expected. It was also found that none of the characteristics can produce according to the pre-defined specifications. These results are normal as this scenario is seen in many real-life processes. However, in order to improve performance and avoid un- desirable situations, we suggest implementing corrective actions. On the other hand, the results achieved in Phase II were not as expected, as they were compromised by the relation- ship between the data series from both phases. Therefore, it was not possible to draw specific conclusions about the behavior of the process in this phase, nor to compare performance be- tween the univariate and multivariate study as was intended.
Over the years, quality has been gaining prominence in many organizations, since there are greater demands from customers who want good products at reasonably low prices. It's an investment that generates an important return, not through profit directly, but through the prevention of errors, for example, which in the long term translates into sustainable growth for companies. The main aim of this work is to drive continuous improvement in the production process through the practice of statistical process control. Monitoring the quality of production sys- tems is fundamental, but reality often diverges from theory and industrial processes turn out to be much more complex than idealized. For this reason, the assumptions for applying statis- tical techniques, such as traditional control charts, are sometimes disrespected, which leads to an increase in the number of false alarms and, consequently, to erroneous conclusions about the process. In these situations, in order to correctly understand the behavior of the process, it is important to study the process in depth, using different statistical tools, so that the data can be properly processed. The proposed methodology was implemented in a beverage industry, ECM - Empresa de Cer- vejas da Madeira, to understand and validate its applicability. The study focused on Coral Branca beer, in 5 characteristics: Alcohol, Original Extract, Foam Stability, Cloudiness and Bitterness. This methodology is divided into two phases: Phase I aims to check whether the process is under statistical control and if has the capacity to produce according to the established re- quirements, and then Phase II, where the process is constantly monitored over time. To this end, univariate and multivariate control charts were applied so that the results could be used to make a comparative analysis of the performance of both charts. The application of the different control charts in Phase I showed that the process varies beyond what is considered normal or expected. It was also found that none of the characteristics can produce according to the pre-defined specifications. These results are normal as this scenario is seen in many real-life processes. However, in order to improve performance and avoid un- desirable situations, we suggest implementing corrective actions. On the other hand, the results achieved in Phase II were not as expected, as they were compromised by the relation- ship between the data series from both phases. Therefore, it was not possible to draw specific conclusions about the behavior of the process in this phase, nor to compare performance be- tween the univariate and multivariate study as was intended.
Descrição
Palavras-chave
Controlo Estatístico do Processo Cartas de Controlo Variabilidade Independência Estudo Multivariado Indústria Cervejeira