varga t. – pokorádi l., possibilities of using fuzzy logics in the

10
Varga T. – Pokorádi L., Possibilities of Using Fuzzy Logics in the Quality Planning in the Automobile Industry , PROCEEDINGS of the 12TH MINI CONF. ON VEHICLE SYSTEM DYNAMICS, IDENTIFICATION AND ANOMALIES, BUDAPEST, NOV. 8-10. 2010 ISBN: 978 963 313 058 2, p. 311 – 318. POSTPRINT

Upload: others

Post on 13-Mar-2022

6 views

Category:

Documents


0 download

TRANSCRIPT

Varga T. – Pokorádi L.,

Possibilities of Using Fuzzy Logics in the Quality Planning in the Automobile Industry ,

PROCEEDINGS of the 12TH MINI CONF. ON VEHICLE SYSTEM DYNAMICS, IDENTIFICATION AND ANOMALIES, BUDAPEST, NOV. 8-10. 2010

ISBN: 978 963 313 058 2, p. 311 – 318.

POSTPRINT

POSSIBILITIES OF USE OF FUZZY LOGIC IN THE QUALITY PLANNING IN THE AUTOMOTIVE INDUSTRY

Tamás VARGA* and László POKORÁDI� *National Instruments Kft �University of Debrecen

H-4028 Debrecen, Ótemető u. 2–4. [email protected], [email protected]

Received: November 10, 2010

ABSTRACT The quality planning in the automotive industry is a really complex system, which contains a lot of elements. When we would like to modeling this system, the easiest way is to modeling the certain parts. The quality planning maybe is the most elaborated in automotive industry, it is known as APQP. The process is epitomized well in the APQP reference manual, which is a very good guideline that is why we will collect the elements of the system from this manual, what we can model with fuzzy logic. The main aim of paper is to examine, what elements of the quality planning in the automotive industry can be modeling by fuzzy logic follow up the similarities between fuzzy logic and quality planning.

Keywords: APQP, fuzzy logic, FMEA, quality function deployment, design of experiment

1. INTRODUCTION

Nowadays only a few hundred automotive suppliers have ISO TS 16949 certification, some of them in Hungary. In the next following years the growth probably will be huge, as the “forbear” QS 9000 certification was over fourteen thousand all over the world. It shows well the importance of quality planning, and it will increase more and more. The success of quality planning depends on the success of the certain elements like QFD, FMEA or DoE. In this paper we will discover the advantage of using fuzzy logic in the certain APQP tools. Fuzzy logic is one of the multivalue semantics. It has been examined by many scientists in their studies, so it has a huge literature nowadays, but the foundations were estabilished by Lotfi A. Zadeh’s article [13] in 1965. He examined the smudgy border of true range of notions in vulgar tongue from mathematical viewpoint. He has given the “fuzzy” name to this field. He assigned a value from [0; 1] closed interval to each logic statement during the model making. Originally he defined only the concept of the fuzzy sets and the characteristic functions of these sets, the so called fuzzy functions. Retter in his work [10] gives us a great explanation to present the key of fuzzy success, he shows the advantage, what we would like to follow up in this paper. Pokorádi in his book [5] and in his articles [6] [7] [8] examines the possibilities of mathematical model making about systems, one of the possibilities from the presented is fuzzy modelling. In the chapter regarding this topic, it is shown how to use it when proceed a fuzzy FMEA. This is very important for us, because as we will see FMEA has a huge importance in success efficiency of APQP projects. Portik has an interesting article [9], where he examines the combination of fuzzy with management. The authors investigated possibilities of use of fuzzy set theory in the quality planning [11]; [12]. The rest of the paper is organized as follows: Section 2. words the structure of APQP. The best start to learn more about this project management tool is to review the official standard, so in the second chapter we look through the APQP guideline [1]. Section 3. presents the

essence and the build-up of fuzzy logic. Section 4. shows the APQP tools which use people’s experiences and opinions as input, so give a very good base to develop with fuzzy. We can see examples for each tool as well. In section 5., the summary of this work and authors’ future work are given.

2. STRUCTURE OF APQP

APQP can be used as a guideline during the development/design phase; furthermore it is a standardized frame to ensure the communication of the results between customer and suppliers.

Fig. 1. Product Quality Planning Timing Chart (source: [1])

APQP guideline determines three main phase: - Development - Industrialization - Product Launch These chapters contain 23 topics, all of them are recommended to go through before mass production. Some of these topics: durableness, test planning, specifications, quality inspection requirements, specification of packaging and the process capacity, testing of finished products or training plan for operators. The main questions of APQP: quality planning, measurement of customer satisfaction and transformation customer demands into engineering parameters. The guideline builds up from five chapters: plan and define program; product design and development; process design and development; product and process validation; feedback, assessment and corrective action. Central elements: learn customer demands, proactive feedbacks and corrective actions, planning within the process capability, failure mode and

effect analysis and mitigation, review and validation, check plans, control of special and critical factors. In Fig. 1., the timing chart of APQP can be seen.

3. FUZZY LOGIC

The real world and our brain are extra imprecise, smudged and this cause it is really complicated. Two concepts were worked out to describe this world. First concept is vulgar tongue made by Neanderthal man. It informs us of highly complex facts on really simple way based on general accepted agreements. We are talking about “nice weather”, “successful sport event”, “average speed” with uncertain, smudged character. The imprecision of vulgar tongue is necessary; otherwise we should have to define dozens of definitions continuously. On that way our verbal communication would be unimaginable and untreatable.

Fig. 2. Fuzzy process (Source [11])

The other solution was developed by mathematicians in an era, when precision was the most important fact. In case of this idea, every item, set, function, parameter and everything must be well defined. It means so much definition and condition which caused the mapping of the real world was untreatable. This is the so called hard mathematics, where we have to idealize, simplify and abstract the input parameters and the relationships to make a practicable system. The idealisation, simplification and abstraction cause an inaccurate, inflexible mathematical model. Here comes Zadeh with fuzzy logic and turn back to the first concept to develop the second, which means, he presented a mathematical concept, which can use imprecise, smudged expressions as inputs for mathematical relations and models without simplification. He laid down the fundamentals of precise logic of imprecision. It means, the opinion of experts, operators, engineers can be describable on exact mathematical way. This is the only method what is able to transform the human knowledge and expertise to mathematical knowledge. Let’s see the figure below, which present the working of the system very well.

3.1 Fuzzyfication

The first step is the fuzzyfication, which means the upset of the system with concrete input values. During this step we assigned fuzzy membership value to each typical parameter of the examined system. To describe the inaccuracy and uncertainty of inputs we can apply similar

definitions like equation (1), figure 3.

( )

≤<<−≤≤<<−

x9if05x4ifx54x3if13x2if2x

2xif0

B . (1)

BOOLE set Fuzzy set

Fig. 3. Comparison of Boole and fuzzy sets based on equation (1) (source: [5])

At the first step we have to determine the categories applied in the modelling process and the relationship functions belong to the categories. At the beginning let’s examine the main influential factors. For example in case of the general risk assessment, the risk level is influenced by occurrence of the event and the weight of the possible loss caused by the event. In case of FMEA the influential factors are the occurrence, the severity of consequent and the detectability level of the failure in contrast with general risk assessment. It is really important to choose adequate number of categories, because the mapping will be better and more realistic if we increase the number of categories, but the examination will be more complicated and the possibility of misunderstanding among the expert will decrease. Let’s see the determination of membership function. The declaration of functions for the categories can happened on different ways. The membership function µ(x;A) is present the membership level of x parameter to set A. An important question is the definition of the scale. It is practical to choose between 0 – 10, 1 – 10, 0 – 100, 1 – 100, these ranges will make easier the comparison of the examined items.

3.2 Inference

In the inference phase we have to make logical rules based on the chosen categories, so we have to determine the rule base regarding our fuzzy model.

3.3 Composition

In the summary step we link the values differ from zero – what we got from explanation – using either of fuzzy operation. Fuzzy operations are presented in table 2.

Table 1. Comparison of Fuzzy and Boole - algebraic operations (source:[11])

We will get a fuzzy set, this will be the primary conclusion, that is why we need for the last step, the defuzzyfication.

3.4 Defuzzyfication

In the last step, we choose the value, which is mainly typical for the examined system based on the fuzzy conclusion we got. The meaning of fuzzy set can be different depends on the type of application, hence we can choose from different defuzzification manners to get the proper result. Optional manners can be the Centre of Gravity (COG), Centre of Area (COA) and weighted average of maximums manner.

4. FUZZY LOGIC IN QUALITY PLANNING

Fuzzy logic is the tool, what makes the certain important elements of quality planning to modelable. The mathematical model of the whole process makes basis of scientific examinations. Let us go through the APQP guideline to find connection among the certain parts and fuzzy expert system. The first chapter starts with the so called “Voice of Costumer”. It is used in many fields of the industry, its’ better known name is QFD (Quality Function Deployment). It means the transformation of customer demands into engineering parameters. This is a very useful analytical device in design and development phase. For success of a company in our globalized world nowadays it is necessary to be with products on markets of different countries. When we would like to survey the demand of inhabitants of different countries, we have to consider the linguistic differences and imprecision. We have to calculate, that particular expressions maybe have different meanings in case of different ethnical groups. Fuzzy logic helps us to make handleable this problem. Kahraman et al. shows a really good example in their article [3] for the combination of QFD and fuzzy logic. They demonstrate an example, where they survey the demand of customers and transform it into engineering parameters in case of a Turkish plastic door and window company using fuzzy logic. Three kind of QFD is known depend on the input parameters. The basic input can be market research, when surveys, customer interviews, quality and reliability studies, competitor product studies give the data. In the second case we work up historical data, when best practices, warranty reports, problem solving reports, customer claims give the input. In the third case experience of an expert team provide basic information to proceed QFD. This time lawful regulations, aggravations, vendors, operators (who produce the affected product) opinion, inside evaluations and professional best practices are worked up. In the second chapter of APQP guideline, design failure mode and effect analysis (DFMEA) is mentioned as a tool, which can help to understand the design problems the designers have to face with. We can make a realistic and flexible mathematical model about FMEA as well with using fuzzy logic, moreover we can find further advantages. The conventional FMEA

consider the occurrence, the severity and the detection of a possible failure in the lifetime of the product. The basic inputs are given by opinion of an expert team regarding the probable failures and its effect. As in case of this tool we have to summarize opinions of people (like QFD), so we have to face with the linguistic and conceptual differences, what we could see above. Based on it the authors think we can develop FMEA with combining fuzzy logic. In the third chapter another kind of FMEA is shown, which applied in case of assembly tasks, the so called process FMEA (PFMEA). Of core in this case we can apply a same method like DFMEA and we can reach the same advantages as well. FMEA is not just a developing and design tool but a monitoring manner, so the quality of the product and process can be developed any further by the systematic analysis of occurred failures during the implemented process. We could see, proceeding an FMEA can be happened with different aims and on different ways and it can be developed any more based on fuzzy logic. If we review APQP guideline, we can feel fuzzy FMEA cover the biggest part of a quality planning project, which means it has serious effect for the success of the whole project, so the authentic reflection of the real situation is one of the most important things to consider. K. Xu. at al. has an article [4], in which they point out, that the interdependencies among the various failure modes with uncertain and imprecise information are very difficult to be incorporated for failure analysis. This means, that the validity of the result of conventional FMEA maybe questionable. The authors are performing a conventional and a fuzzy-logic-based FMEA of a diesel engine’s turbocharger system. There are three main advantages for employing fuzzy. First, all FMEA related information is recorded in natural language. It is easy and plausible for fuzzy logic to deal with them because the basis for fuzzy logic can be built on top of the experience of experts. The result will be more realistic and flexible reflection of the real situation. The second thing is that fuzzy logic allows imprecise data to be used as an input. Therefore it can easily treat many states of components and system and other fuzzy information included in FMEA. The interdependencies among various failure modes and effects can be explored. Finally, the expert assessment system – presented by the authors of the article [4] – fully incorporates engineers’ knowledge and expertise in the FMEA analysis and substantial cost savings can thus be realised. There is another field can be developed with fuzzy, this is design of experiments (DoE). This is a standard method, where we change the input parameters of the process between determined limits towards to learn how it does affect for the outputs. It plays significant role proceeding quality systems in the industrial practice. Johanyák shows a good example in his article [2] how to model experiment results based on an adaptive neuro-fuzzy system, which differs from the conventional manner. The main advantage of using fuzzy logic in DoE is, that the mathematical model of the examined process in not necessary to have. We can make the fuzzy model of the process with using different learner algorithms. The biggest advantage of the presented system is that it can make a fuzzy structure, easily interpreted by humans after a short learning period. We examined the possibility to combine of the certain elements of APQP with fuzzy. As APQP is a project management tool, it can be interesting at this point if we examine the advantages of using fuzzy logic in management (in general).

5. SUMMARY, FUTURE WORK

Nowadays only a few hundred automotive suppliers have ISO TS 16949 certification, some of them in Hungary. In the next following years the growth probably will be huge, as the “forbear” QS 9000 certification was over fourteen thousand all over the world. It shows

well the importance of quality planning, and it will increase more and more. The success of quality planning depends on the success of the certain elements like QFD, FMEA or DoE. In this state-of-art paper we have seen that the validity of the result of conventional FMEA maybe questionable, because all FMEA related information is recorded in natural language and the hard mathematics can cope with it only using idealisation, simplification and abstraction, which cause a not completely realistic and flexible reflection. The other important factor that the interdependencies among various failure modes and effects can not be explored without fuzzy. This is true for QFD and DoE as well as the basis for them is people’s experiences and opinions. The aim of the author of this paper is learning other quality planning methods from other field of industry (like aeroplane, aerospace industry) and develop the method with soft mathematical tools to make a more realistic reflection to reach better quality.

6. REFERENCES

[1] Advanced Product Quality Planning (APQP) and Control Plan, Reference Manual, 2008. pp. 107.

[2] Johanyák, Zs. Cs. – Kovács, Sz.: Adaptation of neuro-fuzzy methods in the experiment methodology, Contribution of GAMF, Kecskemét, XX: 2005, p. 37-48. (in Hungarian)

[3] Kahraman, C.- Ertaj , T. - Büyüközkan, G.: A fuzzy optimization model for QFD planning process using analytic network approach, European Journal of Operational Research 171, 2006, p. 390-441.

[4] K. Xu, L. C. Tang - M. Xie - S. L. Ho - M. L. Zhu.: Fuzzy Assessment of FMEA for Engine System, Reliability Engineering and System Safety 75, Elsevier Science Ltd., 2002, p. 17-29

[5] Pokorádi, L.: Systems and Processes Modeling, Campus Kiadó, Debrecen, 2008, pp. 242. (in Hungarian).

[6] Pokorádi, L.: Fuzzy Techniques in the Aircraft Engineering, Proceedings on the 7th Mini Conference on Vehicle System Dynamics, Identification and Anomalies, VSDIA 2000, Budapest 2000., p. 443–448.

[7] Pokorádi, L.: Diagnostic-Connected Decisions based on Expert Knowledge, Proceeding of 8th Mini Conference on Vehicle System Dynamics Identification and Anomalies, VSDIA 2002, Budapest, 2002., p. 681–687.

[8] Pokorádi, L.: Anomaly Effect Analysis based upon Estimate of the Experts, Proceeding of 9th Mini Conference on Vehicle System Dynamics Identification and Anomalies, VSDIA 2004, Budapest, 2004., p. 529–535.

[9] Portik, T. – Pokorádi, L.: Possibility of use of fuzzy logic in management, 16th Building Services, Mechanical and Building Industry days" International Conference, 14-15 October 2010, Debrecen, Hungary, p. 353-360.

[10] Retter, Gy.: Fuzzy, neurális genetikus, kaotikus rendszerek, Akadémiai Kiadó, Budapest, 2006

[11] Varga, T.: Possibilities of Use of Fuzzy Logic in the Quality Planning, Debreceni Műszaki Közlemények 2010/1, Debrecen, Hungary, p. 43-51. (in Hungarian)

[12] Varga, T. – Pokorádi, L.: Quality Planning Methods, Development Possibilities of Risk Management Tools Based on Fuzzy Expert System, 16th Building Services, Mechanical and Building Industry days" International Conference, 14-15 October 2010, Debrecen, Hungary, p. 361-366.

[13] Zadeh, L.: A. Fuzzy Sets, Information and Control, 8 (1965), p. 338-353.