|
1.INTRODUCTIONWith the increasingly fierce market competition of electromechanical products, requirement has increasingly become the premise and guidance of innovative design of electromechanical products1. The traditional methods of requirement mining and analysis include analytic hierarchy process2, 3, questionnaire analysis based on Kano model4, 5, etc. The traditional requirement analysis method can help designers analyze customer requirement more accurately. However, this method usually depends on manual consultation, questionnaire survey and other activities. The research cost is high, and it is easy to produce information deviation and insufficient reliability. It is difficult to apply to the massive outbreak of requirement data. With the advent of the data age, many data-driven requirement analysis methods have been proposed, such as consumer requirement data quality evaluation, customer emotion polarity and its corresponding opinion target analysis, and automatic ranking of customer requirement importance6. Yao Y proposes a three-stage method to identify product feature names from customer opinion data. In the first stage, the brown clustering algorithm is applied to obtain word clusters with similar meanings, and the changes of brand names are captured by language rules. In the second stage, the method based on conditional random fields (CRFs) is used to analyze whether a word really refers to a specific product model name. In the third stage, rule-based name normalization is used to map names to their formal names7, 8. Zhou F uses fast-text technology to obtain comments containing useful information from Internet product reviews, extracts various topics related to customer needs through a topic modeling technology, and finally predicts the emotion category and intensity of consumers in comments through rule-based emotion analysis9. Polpinij J uses support vector machine and other technologies to classify customer demand text data, extract the characteristics of customer demand text data, and then analyze the emotional polarity10. With the development of e-commerce, the amount of personalized requirement data in the form of customer comments on internet platforms has exploded. Because the customers are generally not professionals, the expressed requirement has the characteristics of fuzziness and incompleteness. On the one hand, the storage cost and processing cost of massive low-quality data are high, and the acquisition cost of massive high-quality data are high; On the other hand, there are many missing values in the requirement data, and it is difficult to accurately quantify the data, categories and attributes. Therefore, knowledge aided methods and tools are needed to assist designers and customers to quickly clarify requirements. The method of knowledge aided requirement analysis can reduce the burden of designers dealing with massive personalized customer requirement data, so that designers can focus more on the links of innovative design. At present, knowledge aided requirement analysis methods include knowledge retrieval based on keyword and file name matching11, rule-based reasoning12 and case-based reasoning13. The method based on keyword and file name matching is simple and easy to implement, but the knowledge granularity is large, which is difficult to be directly used by designers. The method of rule-based reasoning has simple knowledge representation and strong reasoning ability. However, with the expansion of knowledge scale, its reasoning efficiency will decline sharply. The case-based reasoning method can solve new problems by using the past requirement analysis cases, and solve new problems by retrieving the most similar cases. However, the interpretability of knowledge acquisition in this method is weak. We propose a method of electromechanical product requirement identification and disassembly based on domain knowledge network. This method can identify and disassemble the requirement text expressed in the unstructured form of natural language, and obtain the keywords that can accurately and fully represent the meaning of customer requirement. The framework of this method is shown in Figure 1. Firstly, we construct a domain knowledge network based on ontology, and then reuse knowledge in the process of customers requirement analysis. Because ontology has the characteristics of standardized expression and accurate description of the relationship between entities, this method can describe knowledge at the level of entity granularity, and has advantages in knowledge reasoning. Finally, based on this method, we develop a requirement identification and disassembly tool for electromechanical products to assist designers and customer to quickly clarify requirement. 2.CONSTRUCTION METHOD OF DOMAIN KNOWLEDGE NETWORK BASED ON PATENTIn order to support the requirement identification and disassembly method of electromechanical products based on domain knowledge network, we propose an automatic construction method of domain knowledge network based on patent. Because the requirement description of electromechanical products has the characteristics of semi professionalism and concealment. We divide the requirement elements into functional elements and structural elements based on axiomatic design theory, and expresses the functional elements into function-operation and function-object based on behavior theory. The process of this method is shown in Figure 1.
3.REQUIREMENT IDENTIFICATION AND DISASSEMBLY METHOD OF ELECTROMECHANICAL PRODUCTS BASED ON DOMAIN KNOWLEDGE NETWORKNow we can realize the requirement identification and disassembly method of electromechanical products based on the domain knowledge network. We divided these keywords into Core Requirement Keyword (CRK), Important Requirement Keyword (IRK) and General Requirement Keyword (GRK). The CRK is the explicit requirement for product functions, structures and scene clearly expressed by customer in the requirement text. The IRK are the standardized expression of the CRK based on the domain knowledge network, which is the result of matching the CRK with the relevant entities in the domain knowledge network. GRK are the results of mining implicit requirement that are not clearly expressed by customer based on IRK and domain knowledge network. They can accurately and fully express the meaning of the customer requirement. As shown in Figure 1, the specific process of this method is:
4.APPLICATION DEVELOPMENTBased on the requirement identification and disassembly technology of electromechanical products based on domain knowledge network, we develop the requirement identification and disassembly tool of electromechanical products based on domain knowledge network. The framework of the tool is shown in Figure 3. 5.APPLICATION EXAMPLETaking a real electromechanical product requirement description text as an example, we introduce the process of electromechanical products requirement identification and disassembly tool based on domain knowledge network, so as to verify the feasibility of the requirement identification and disassembly method based on domain knowledge network. The process and results are as follows:
Table 1.Mining and recommendation results of IRK.
Table 2.GRK mining and recommendation results of “pressure relief valve”.
Table 3.GRK mining and recommendation results of “pressure relief”.
6.CONCLUSIONS AND PROSPECTSThis research mainly focuses on the problem of requirement analysis in the field of electromechanical products, and puts forward the method of requirement identification and disassembly of electromechanical products based on domain knowledge network. There are three main contributions: (1) We propose a patent based automatic construction method of electromechanical domain knowledge network, which can automatically complete the process of obtaining relevant entities and relationships from patents, and finally build a domain knowledge network that can support the requirement identification and disassembly methods of electromechanical products; (2) We propose a method of requirement identification and disassembly of electromechanical products based on domain knowledge network, which can obtain keywords that can accurately and fully reflect the customer requirement, and help designers effectively mine and analyze the customer requirement. (3) We develop a requirement identification and disassembly tool for electromechanical products based on domain knowledge network. The research still has deficiencies in some aspects, which needs to be expanded and deepened in the follow-up research. (1) We do not fully analyze the characteristics of the scene in the customer requirement text and fails to clearly define the scope and category of the scene. In the process of customer requirement identification and disassembly, the scene elements contained in the requirement are not well utilized. (2) The knowledge network in the method of identification and disassembly of mechanical and electrical products based on knowledge network is lack of scalability. In practical engineering application, knowledge has not only static independent knowledge fragments, but also dynamic continuous knowledge. Knowledge has an evolution process, and practical engineering will continue to need new knowledge. Therefore, the knowledge network in practical engineering application should be expandable in order to incorporate new knowledge in time. ACKNOWLEDGMENTThe research was supported by a National Key Research and Development Project (No. 2018YFB1700802). REFERENCESLi, X. Z., Research on Customer Value Oriented Conceptual Design Method of Intelligent Products, Shanghai Jiaotong University, (2017). Google Scholar
Saaty, T. L., The Analytic Hierarchy Process, McGraw-Hill, (1980). Google Scholar
Nepal, B., Yadav, O. P. and Murat, A.,
“A Fuzzy-AHP approach to prioritization of cs attributes in target planning for automotive product development,”
Expert Syst. Appl, 37
(10), 6775
–6786
(2010). https://doi.org/10.1016/j.eswa.2010.03.048 Google Scholar
Matzler, K. and Hinterhuber, H. H.,
“How to make product development projects more successful by integrating Kano’s model of customer satisfaction into quality function deployment,”
Technovation, 18
(1), 25
–38
(1998). https://doi.org/10.1016/S0166-4972(97)00072-2 Google Scholar
Shen, X. X., Tan, K. C. and Xie, M.,
“An Integrated approach to innovative product development using Kano’s model and QFD,”
Eur. J. Innovation Manage, 2 91
–99
(2000). https://doi.org/10.1108/14601060010298435 Google Scholar
Jin, J., Liu, Y., Ji, P., et al.,
“Review on recent advances in information mining from big consumer opinion data for product design,”
Journal of Computing and Information Science in Engineering, 19
(1), 010801
(2019). https://doi.org/10.1115/1.4041087 Google Scholar
Yao, Y. and Sun, A,
“Product name recognition and normalization in internet forums,”
in SIGIR 14,
(2014). https://doi.org/10.32657/10356/61814 Google Scholar
Yao, Y. and Sun, A,
“Mobile phone name extraction from internet forums: A semi-supervised approach,”
World Wide Web, 19
(5), 783
–805
(2015). https://doi.org/10.1007/s11280-015-0361-1 Google Scholar
Zhou, F., Ayoub, J., Xu, Q., et al.,
“A machine learning approach to customer needs analysis for product ecosystems,”
Journal of Mechanical Design, 142
(1),
(2020). https://doi.org/10.1115/1.4044435 Google Scholar
Polpinij, J. and Ghose, A. K.,
“An ontology-based sentiment classification methodology for online consumer reviews,”
in IEEE/WIC/ACM Inter. Conf. on Web Intelligence and Intelligent Agent Technology,
518
–24
(2008). Google Scholar
Iyer, N., Jayanti, S., Lou, K., Kalyanaraman, Y. and Ramani, K.,
“Shape-based searching for product lifecycle applications,”
Computer-Aided Design, 37
(13), 1435
–46
(2008). https://doi.org/10.1016/j.cad.2005.02.011 Google Scholar
Zhang, X. T., Peng, G. L., Hou, X. and Zhuang, T.,
“A knowledge reuse-based computer-aided fixture design framework,”
Assembly Automation, 34
(2), 169
–181
(2014). https://doi.org/10.1108/AA-04-2013-033 Google Scholar
Reyes, E. R., Négny, S., Robles, G. C. and Le Lann, J. M.,
“Improvement of online adaptation knowledge acquisition and reuse in case-based reasoning: Application to process engineering design,”
Engineering Applications of Artificial Intelligence, 41 1
–16
(2015). https://doi.org/10.1016/j.engappai.2015.01.015 Google Scholar
|