Application of a hybrid case-based reasoning approach in electroplating industry
Introduction
In electroplating industry, the cost of electroplating component highly depends on the coating weight of precious metal (e.g. gold, platinum, silver, etc.) that sticks on the component surface. As the electroplating market is vigorously competitive, the price quotation that customers would accept drops continuously. In order to survive in such an industry, a tight cost control approach should be examined. Failing to do so means failure to make profit or may even result in losing money. In this connection, electroplating company demands more accurate coating weight from salesperson in order to calculate a competitive quotation. However, it is not easy to estimate this coating weight. Many factors like surface area, texture, and structure of the component play to vary its value. Apart from the coating weight accuracy, responsiveness is also a critical factor that affects the success of a business transaction. Salesperson may determine the coating weight accurately by trial, but customer demands salespersons replying in a timeliness manner as the entire supply chain in the watch industry is compressing continuously. Salesperson can evaluate the coating weight only by their experience and perception. This can easily lead to wrong decision-making, especially from inexperienced salesperson. Finally, accuracy of laboratory apparatus is another uncertainty in quotation. As thickness and weight of the coating are extremely small in value (up to micron and 10−3 g), considerable measurement error is difficult to avoid even though laboratory staffs always try hard to act patiently on all of their procedures. All of these uncertainties can lead to tremendous error in quotation price. Due to this error, cost for handling precious metal is hugely increased, as material planner needs to keep more safety stock for the vague consumption of precious metal. Thus, improving the accuracy of coating weight can directly contribute to better control of cost management and raw material inventory management.
To solve this problem, a systematic approach incorporates with Artificial Intelligence (AI) is used to filter out uncertainties and to recommend a reasonable and reliable coating weight for salesperson. The methodology should be capable enough to interpret factors that affect the coating weight of each component and summarize the importance of these factors to get a balance for the quotation price. All these analyzing procedures depend highly on experience and perception of salesperson.
From the numerous AI concepts in the past, Case-Based Reasoning (CBR) is used for this system. CBR is a good concept to transform the evaluating experience from professionals into a computer-assessment method to make the evaluation process more convenient and efficient (Chang, Cheng, & Su, 2004). It can capture all ‘memory’ of human being without losing them due to lapse of time and carelessness. Knowledge can be preserved in such system when experienced staff left his position. At the same time, knowledge can also be simplified so that new staff can make a comparatively reasonable decision by inputting some easy-to-assess parameters or criteria into the system. On the other hand, such system acquires the advantages of multi-attribute decision-making transform from human beings. The main difference between human being and mathematical system is that human being tends to classify present cases into groups defined by its experience, whereas mathematical system tries to develop a trend, or establish a relationship between variables. If the previous records are accurate and the incoming new cases are always in similar pattern, mathematical approach will be much preferable as it can generate a more accurate result than human being perception. However, in the coating weight evaluation problem, as numerous attributes (sometimes more than five) should be considered, it is very difficult and complicated to establish a relationship or generate a formula based on these variables. Even though the relationship can be verified, it may need a fundamental change once special new cases are put into the system. Otherwise, the system will give very inaccurate recommendation for cases related to new issues. Therefore, the flexibility of mathematical system is restricted by this approach. It is not recommended to solve this problem by relying on mathematical model only.
Data processing procedure is the soul of case-based reasoning system and directs it to succeed. Without a good information flow, the system would misunderstand how to store the cases and make use of them for decision-making. In this research, Fuzzy Logic (FL) and Rule-Based Reasoning (RBR) concepts are merged into CBR system. FL is a formal tool eminently suited for solving problems with imprecision inherent in empirical data (Entemann, 2002). For some uncertainties, such as apparatus errors, it is useful to apply FL to tackle them. RBR is a natural knowledge representation, in the form of ‘If…Then…’ structure (Lin, Tseng, & Tsai, 2003). It makes up of inference engine and assertion, which will be employed for interpreting sets of facts and rules.
This paper includes six main sections. Section 2 summarizes the CBR concept that has been developed by previous researchers. Section 3 describes the proposed methodology and its innovative concept from other CBR approaches. Section 4 applies the developed methodology into the preset problem—electroplating component quotation problem. Section 5 shows the results of the industrial case study with detailed discussion. Section 6 concludes the research outcome.
Section snippets
Literature review
CBR has already been applied in numerous areas. Harrision (1997) found that CBR has broadly been applied in Diagnosis, Help Desk, Assessment, Decision Support and Design. Spalazzi, 2001, Carswell et al., 2002 have similar perspective, they further extended the application domain to software reuse, Logistics, Processing Planning, Navigation Planning, Imagery, Image Recognition and Segmentation, Sketch-based retrieval of architectural data. It can be observed that researchers have expanded CBR
Specific requirements of the proposed system
Before illustrating the procedure for developing CBR system, it is important to understand the specific requirements of the proposed system as follows:
- 1.
The proposed system is able to deliver numerical values.
- 2.
Highly accurate solution: the value estimated should be up to three decimal places. Too rough estimation is useless for salespersons to set an appropriate price.
- 3.
Effective algorithm for tackling inaccurate attributes to case representation: as the value estimated by the system is in
Industrial case study
The methodology described above is applied to solve quotation problem exists in electroplating industry. The main components that are considered in the system are watch parts, which include band, top ring, case, etc. All problems and their corresponding case representation attributes are first illustrated in the form of component-attribute hierarchy as shown in Fig. 3.
The model number of the component is first input into the system to search for previous record. If same component has been
Results and discussions
The error of the coating weight, which was estimated by system and salesperson, is shown in Fig. 5. Before starting this comparison, the system has already store 130 cases, which is related to the coating weight of the testing component.
From Fig. 5, we can see that the value generated by CBR system are much more accurate than those by the salesperson. The average absolute error taken from these 30 cases is 14.72%, smaller than what the salesperson perform, which is 20.05%. Moreover, there
Conclusion
This research chooses electroplating industry, an area with no AI background, for application of CBR system. The original concept of CBR: making people not to think and remember much (Riesbeck, 1996), is successfully incorporated into the proposed hybrid CBR system. Moreover, RBR and FL concepts are added to the system in order to improve system accuracy and shorten the learning duration. In addition, this research redefines case as an objective selection rule, so that CBR can be interpreted as
References (33)
- et al.
Using case-based reasoning to establish a continuing care information system of discharge planning
Expert Systems with Application
(2004) - et al.
A parallelized indexing method for large-scale case-based reasoning
Expert Systems with Application
(2002) - et al.
Improving accuracy by combining rule-based and case-based reasoning
Artificial Intelligence
(1996) - et al.
Using a hybrid rule-based approach in developing an intelligent tutoring system with knowledge acquisition and update capabilities
Expert Systems with Applications
(2004) - et al.
Predicting information systems outsourcing success using a hierarchical design of case-based reasoning
Expert Systems with Applications
(2004) - et al.
Rough set approach to case-based reasoning application
Expert Systems with Applications
(2004) - et al.
Using case-based reasoning to evaluate supplier environmental management performance
Expert Systems with Applications
(2003) - et al.
Document-based workflow modeling: A case-based reasoning approach
Expert Systems with Applications
(2002) - et al.
Design and implementation of new object-oriented rule-base management system
Expert Systems with Application
(2003) - et al.
Using rule induction for knowledge acquisition: An expert systems approach to evaluating material errors and irregularities
Expert Systems with Applications
(1995)
Case-based reasoning: Foundational issues, methodological variations, and system approaches
AI Communications
Case based reasoning: An overview, IIIA Research Report 96
AI Communications Journal
Fuzzy logic: Misconceptions and clarifications
Artificial Intelligence Review
Cited by (66)
Artificial intelligence in industrial design: A semi-automated literature survey
2022, Engineering Applications of Artificial IntelligenceLong-term knowledge evolution modeling for empirical engineering knowledge
2017, Advanced Engineering InformaticsCitation Excerpt :As the first and foremost step in EEK management, a proper representation of EEK can largely facilitate the subsequent EEK acquisition, accumulation and reuse. Chan [25] firstly proposed that empirical engineering knowledge could be characterized by some discrete and standard attributes. Based on this idea, several scholars concentrated on EEK representation [14,15,26–30], and a concise summary of EEK representation in their works is listed in Table 1.
An ontology-based personalized retrieval model using case base reasoning
2014, Procedia Computer ScienceAn integrated case-based reasoning approach for personalized itinerary search in multimodal transportation systems
2013, Transportation Research Part C: Emerging TechnologiesCitation Excerpt :Case-Based Reasoning (CBR) was first proposed by Schank and Abelson during the late 1970s (Shank and Albreson, 1977). The CBR technique is classified as one of the learning approaches for Artificial Intelligence (AI) (Chan, 2005). CBR is used to solve a new problem by remembering a previous similar situation and by reusing information and knowledge about that situation (Aamodt and Plaza, 1994).
A new adaptation method based on adaptability under k-nearest neighbors for case adaptation in case-based design
2012, Expert Systems with ApplicationsCitation Excerpt :Therefore, case retrieval (CR) with similarity measurement (SM) and case adaptation (CA) after CR stage are the two most important processes in CBD (Aamodt & Plaza, 1994; Finnie & Sun, 2003). So far, SM has already been studied widely, such as distance SM (Bryant, 1997; Cheng, Jia, Gao, Wu, & Wang, 2008; Elhadi, 2000; Kwong & Tam, 2002; Yu & Liu, 2006a), adapted distance SM (Bajoa et al., 2008; Farah, Boulila, Ettabaâ, Solaiman, & Ahmed, 2008; Fyson & Toll, 2008; Montague, Martin, & O’Malley, 2008), hybrid SM (Chang, Liu, Lin, Fan, & Ng, 2009; Chiu, Chang, & Chiu, 2003; Han, Shen, & Yang, 2002; Li, Sun, & Sun, 2009; Robles, Negny, & Le Lann, 2008), fuzzy SM (Chan, 2005; Chang, Liu, & Lai, 2008; Khanum & Shafiq, 2008; Li & Ho, 2009; Mendes & Mosley, 2008; Wu, Lo, & Hsu, 2008), and adapted fuzzy SM (Qi, Hu, Peng, Wang, & Zhang, 2009, 2011; Stéphane & Marc, 2008; Zbigniew, Maria, Stefan, & Marcin, 2005). Although abundant literatures focusing on CR with SM study, Biermans et al. (2008) pointed out that many case-based systems are so-called retrieval-only systems or act primarily as retrieval and reuse systems.
Case reuse based product fuzzy configuration
2011, Advanced Engineering InformaticsCitation Excerpt :Gu et al. [18] concluded these researches as three categories: case-biased, query-biased and equally-biased similarity calculation methods. According to the uncertainty issue of valuating the case attributes, some studies [19–21] proposed a fuzzy-CBR approach. By using natural language form or fuzzy number to represent the linguistic variable, the similarity metrics were calculated and the case-retrieving mechanism was implemented.