Bayesian System in Manufacturing (Editing Sample)

Bayesian System in Manufacturing
Your name
Institution’s name
Couse number: course name
Instructor’s name
Due date
Abstract
Bayesian networks (BN) provide a sound theoretical framework that assists in decision making regarding complex and uncertain domains. It does so by assembling diverse information in a consistent and coherent framework, which later incorporates the uncertainties inherent in legitimate systems. The networks possess various processes and purposes for the production system. It has been implemented for the classification of defects, root cause analysis, estimation of reliability of the manufacturing system, and many purposes. Practical applications of BN involve conventional processes like welding, casting, machining. Furthermore, micromachining equally applies to BN. The absolute advantage of Bayesian networks over other statistical learning processes stays flexible to incorporate domain knowledge. Moreover, it prepares the mathematical model that has been noticed in all the literature reviews. Bayesian networks have been successfully used to assist problem-solving in an extensive range of disciplines, including information technology, modern architecture, pharmacy, engineering, construction, neuroscience, ecosystem, biology, and ecology.
Keywords: Bayesian network, Complexity, Directed acyclic graph, Graphical model, domain, node, probability table.
Decision-making tends to be very challenging, as the information extensively involves uncertainty, incompetence, and unreliability. As probability theory is mathematically sound, it provides an adequate basis for modelling real-life problems. Bayesian networks provide a sound theoretical framework that assists in decision making regarding complex and uncertain domains.Bayesian estimation has been extensively used with neural networks as an effective tool in many industrial problems, incorporating expert's extensive knowledge for designing the system was difficult because of the various limitations experienced.[Holmes, D., & Jain, L. (2008). Introduction to Bayesian Networks. Innovations in Bayesian Networks, 1-5.]
One of the main requirements of using Bayesian networks is to have a large dataset to train the model. However, this shortcoming has also been addressed. All the computations related to Bayesian networks are exhaustive and long. With the aid of diverse available computer applications, we easily tackle his problem. For probability calculations in BN, multiple Computer programs get accesses such as Netica, AutoMod , C++, and much more.[Liu, Y., & Jin, S. (2012). Application of Bayesian networks for diagnostics in the assembly process by considering small measurement data sets. The International Journal of Advanced Manufacturing Technology, 65(9-12), 1229-1237.] [Rodrigues, M., Liu, Y., Bottaci, L., & Rigas, D. (2000). Learning and Diagnosis in Manufacturing Processes through an Executable Bayesian Network. Intelligent Problem Solving. Methodologies and Approaches, 390-396] [Durak, J., Adrian, A., Mrzyglod, B., & Kluska-Nawarecka, S. (2007). BAYESIAN NETWORKSAND FUZZY-LOGIC IN HOT DIP ZINC GALVANIZING DEFECTS DIAGNOSTICS. IFAC Proceedings Volumes, 40(18), 727-732] [Masruroh, N., & Poh, K. (2007). A Bayesian network approach to job-shop rescheduling. 2007 IEEE International Conference on Industrial Engineering and Engineering Management] [Rodrigues, M., Liu, Y., Bottaci, L., & Rigas, D. (2000).]
This review paper comprises of 4 main sections. In the next section, there is a brief discussion of the theoretical background of Bayesian networks. In section 3, we will look at practical applications of Bayesian networks in the production system. Section 4 is the culmination of the previous sections, which articulates the importance of Bayesian Networks.
Theoretical Background
Bayesian networks are a type of probabilistic graphical model that uses Bayesian inference for probability computations. The networks are computationally efficient, and they represent conditional dependence, and causation, with the aid of representing conditional dependence using edges in a directed acrylic graph (DAG). The graph doesn't have a closed cyclic loop and quantitative probabilistic information interprets the graph. Through these relationships, you can correctly conduct inference on the random variables within the graph using things. Below described are the general steps involved in contrasting BNs:[Durante, D., & Dunson, D. (2018). Bayesian Inference and Testing of Group Differences in Brain Networks. Bayesian Analysis, 13(1), 29-58.] [Masruroh, N., & Pooh, K. (2007). A Bayesian network approach to job-shop rescheduling. 2007 IEEE International Conference On Industrial Engineering And Engineering Management] [Durante, D., & Dunson, D. (2018). Bayesian Inference and Testing of Group Differences in Brain Networks. Bayesian Analysis, 13(1), 29-58.]
Figure SEQ Figure \* ARABIC 1 Steps involved in constructing Bayesian network[Nannapaneni, S., Mahadevan, S., & Rachuri, S. (2016). Performance evaluation of a manufacturing process under uncertainty using Bayesian networks. Journal Of Cleaner Production, 113, 947-959]
The nodes in the BN represent different process variables relevant to the system, that can be either discrete or continuous. For applying the BN in manufacturing and production, most of the continuous variables got discretized. Discretization of different levels of select the best level of accuracy. Even different level of discretization can also be performed to select the best level of accuracy. A Bayesian network approach has the following advantages over other approaches:[AbellánNebot, J., Morales-Menéndez, R., Vallejo Guevara, A., & Rodríguez, A. (2006). SURFACE ROUGHNESS AND CUTTING TOOL-WEAR DIAGNOSIS BASED ON BAYESIAN NETWORKS. IFAC Proceedings Volumes, 39(13), 408-413] [Khlybov, O., & Dubinin, I. (2012). Algorithm for Controlling Mechanical Properties of Hot Rolled Steels Using Bayesian Network Model. Materials Science Forum, 706-709, 1444-1447.]
Functionality
With the help of joint probability function, a range of tasks can be performed like prognosis of mechanical properties (direct task) or finding the cause of the defect when the defect is known (inverse task).
Structure
The researcher determines a Bayesian model's structure, where they can easily incorporate their expert knowledge and different technological peculiarity which he/she observes day today with the help of probability distributions. This structure is easy to build with the help of available programs already mentioned.
Flexibility
BNs are phenomenological models, so there are no unphysical parameters or specific co-efficient for incorporation. The specific CPTs set the course of the mode, and the CPTs are either obtained by historical data or with the help of expert knowledge.
The BNs can be classified into 3 main types:[Cai, B., Huang, L., & Xie, M. (2017). Bayesian Networks in Fault Diagnosis. IEEE Transactions on Industrial Informatics, 13(5), 2227-2240.]
BN
Also known as static BN, is most widely used in fault diagnosis related problems and different industrial problems. Static BN is very easy to create and comprehend, but when complex systems are involved, inevitable difficulties arise with static BN.
DBN
Dynamic Bayesian network considers how the system is poised to perform over time. Static BNs have higher accuracy when the system hasn’t gone through various changes. But with time, machining components undergo performance degradation. DBN can easily incorporate that, and that’s what makes it more robust than BN. DBN is nothing but an extension of BN with time-dependent variables.
OOBN
Object oriented approaches have several characteristics, such as encapsulation, inheritance, polymorphism, and modularity. These characteristics encoded into BN to create OOBN. OOBNs mainly have two advantages over BNs and DBNs. First, the OOBN supports a top-down model construction approach. Second, the constructions consist of integrating small and understandable network fragments.[Madsen, A., Søndberg-Jeppesen, N., Sayed, M., Peschl, M., & Lohse, N. (2017). Applying Object-Oriented Bayesian Networks for Smart Diagnosis and Health Monitoring at both Component and Factory Level. Advances in Artificial Intelligence: From Theory to Practice, 132-141.]
Practical applications
Bayesian networks are applied for different processes and purposes in production system. It has been used for classification of defects, root cause analysis, estimation of reliability of manufacturing system and for many purposes. Not only it has been used for conventional processes like welding, casting, machining but it has also been applied for micromachining.[Pernkopf, F. (2004). Detection of surface defects on raw steel blocks using Bayesian network classifiers. Pattern Analysis and Applications, 7(3), 333-342] [Dey, S., & Stori, J. (2005). A Bayesian network approach to root cause diagnosis of process variations. International Journal of Machine Tools and Manufacture, 45(1), 75-91.] [Langseth, H., & Portinale, L. (2007). Bayesian networks in reliability. Reliability Engineering & System Safety, 92(1), 92-108.] [Fujii, H., & Ichikawa, K. (2001). Estimation of weld properties by Bayesian neural network. Welding International, 15(12), 935-939.] [Thomas, P., Suhner, M., Meutelet, B., & Brachotte, G. (2004). Q...