My name is Mamunur Rashid. I have come from Bangladesh. I am working at Bangladesh Institute of Management in Dhaka. The title of my presentation today is “ Decisionmaking in Conceptual Phase of Product Development.” This presentation is related to the doctoral thesis. I have submitted recently.
The agenda is, as follows: General Background, Mathematical Settings, Customer Needs Assessment, Sustainability Assessment, Creativity Assessment, and Discussion and Concluding Remarks.
First, I would like to present the General Background.
Let me introduce you what understand by the phrase called product development. Product development is the study of activities of a product life-cycle in a concurrent manner.
A product development process usually starts by the interaction of internal and external customers. External customers are real customers and the internal customers are the persons directly involved in the product development process (team members). The interaction drives the process to the phase called conceptual phase, wherein , product development team members can decide the key solutions. Key solutions determine more than 80% cost of the product. The next phase is called materialization, wherein product development team members make a detailed design of key solutions. In addition, product development team members can design the manufacturing systems for manufacturing the product. After the materialization, the external customers start to use the product to enjoy the desired satisfaction. After the use, the product is disposed off. Sometimes recycling/down-cycling, landfill, etc. are carried out in this phase.
In conceptual phase, one experiences a lack of knowledge and abundance of choice. A great deal of personal preferences, judgments, imprecisely defined entities, incomplete/vague information underlies this phase. Thus, decisionmaking in conceptual phase is a difficult task to perform.
In my thesis, I would like to emphasize three major issues, namely, 1) Creativity 2) Sustainability and 3) Customer Needs. Internal customers (in other words product development team members) need to exercise something that I can refer to as creativity. Key solutions should be creative ones. How to differentiate a creative-key-solution from a non-creative-key-solution is one of the decisionmaking challenges for internal customers. I will show you some results on this issue later.
The second issue I would like to emphasize is sustainability. Sustainability has earned a great deal of importance because of growing environmental concerns. I will show you later, h ow to deal with sustainability in key solution determination process .
The other issue i would like to emphasize is customer needs. When the internal customers interact with the external customers for the sake of key solution determination, the following issues arise: What is the appropriate customer need model? How to deal with the unknown customer needs? How to classify the key solutions based on customer responses? I will show you later how to answer the above-mentioned issues.
Now, I would like to present the mathematical settings used in this study. As I mentioned before, the decisionmaking in conceptual phase experiences personal preference, judgment, lack of knowledge, and alike. Therefore, for formal computation, I can rely on multi-valued logic and granular information. Thus, in this study I would like to use some mathematical settings based on fuzzy numbers.
Let me introduce the mathematical entity called fuzzy number. A fuzzy number A is defined by a membership function mu A. mu A is a mapping from X to [0,1]. X is a segment of real line and called the universe of discourse. The maximum membership value of mu A is equal to one. There are other conditions that mu A must fulfilled, namely, convexity, continuity, and compactness. The fuzzy number you see in this slide is a fuzzy number called Comfortable (temperature).
Note that: A ( x ) is called the membership value of x with respect to A. A ( x ) is also called the degree of belief of x in terms of A. The Truth-Value ( TV ) of the proposition “ x is A ” is equal to A ( x ).
I can apply an operation called alpha-cut to get a numerical range from a given fuzzy number. This slide shows a range of temperature [20,30] derived from the fuzzy number comfortable using the alpha-cut at alpha equal to 0.5.
I can calculate the expected value of a fuzzy number using a method called centroid method. For the fuzzy number comfortable, the expect value is equal to 25.
This slide shown how to define Support of a fuzzy number. Accordingly, the support of the fuzzy number called comfortable is equal to [15,35].
In my study, I consider linguistic likelihood (fuzzy number) to deal with the imprecision in the probability. I divide the universe of discourse of relative frequency by linguistic likelihoods called most unlikely, quite unlikely, some likely, not sure, some likely, quite likely, and most likely. Other definitions are also possible, as shown in the next slide.
I can even define the linguistic likelihoods by the fuzzy numbers called most unlikely, perhaps unlikely, not sure, perhaps likely, and most likely. I will show you later, the sensitivity of case-1 and case-2 of linguistic likelihoods in customer needs assessment.
Another useful operation on fuzzy number is shown here called range compliance. The range compliance measures the compliance of a numerical range L with respect to a fuzzy number A. Accordingly, the range compliance of a temperature range [10,30] with respect to the fuzzy number comfortable is 0.583.
Range compliance is particularly useful in sustainability analysis because the information is often given by numerical range. This slide shows an example of range compliance of a range [0,5] with respect to fuzzy numbers Very Low, Low, Moderate, High, and Very High, defined in the universe of discourse [0,15]. The summation of range compliance of Very Low and Low is called Desirable Impact. The summation of the range compliance Moderate, High, and Very High is called Undesirable Impact. I can get a zone called good for environment in the plot of UI vs DI. The case shown here comply with the zone called good. Later, this slide will be used as a standard slide for evaluation hard materials. The eco-attribute complying more with very low or low has low impact on the environment. The eco-attribute complying more with moderate, high, or very high has impact on the environment. A procedure is shown to aggregate these positive and negative impacts for all eco-indicators for a given family of materials namely Alumina, Zirconia, Silicon Carbide, Boron Nitride, and Boron Carbide.
This slide shows an example that does not comply with the zone called good.
I have used discrete event simulation to know the unknown answers of external customers. Here, Z is the simulated event which can be either Like, or Must-be, or Neutral, or Live-with, or Dislike. To simulate Z, i can use five rules. The rules use the cumulative probabilities of the possible states of Z. The cumulative probability is calculated from the Probability. The Probability is calculated from the Expected Value of Linguistic Likelihood. The Expected Value of the linguistic likelihood is calculated from the linguistic likelihood of relative frequency. The linguistic counterpart of relative frequency is the linguistic likelihood corresponding to the maximum membership value. Random is a uniformly distributed variable in the interval [0,1].
Now, I would like to explain the information content underlying a given set of propositions. Here i have three propositions, P1, P2, P3, based on three fuzzy numbers, cold, comfortable, hot. Truth-Value of P1, P2, P3 is the membership values, as you can see.
A truth value contains an amount of information content, I , as shown in this slide. The average information content of all propositions is the certainty entropy, CE. CE quantifies the state of knowledge. If CE is equal to 0, knowledge is complete. If CE is equal to one, knowledge is incomplete. In classical logic, CE is always equal to zero.
In this slide I would like to explain the concept of Requirement Entropy RE . To calculate RE first , I would like to introduce a proposition called Requirement Proposition, R . The truth value of R is calculated from the truth values of the given propositions. If the truth value of R is close to the maximum truth value of the given propositions, I can get low Requirement Entropy. Otherwise, I can get high RE .
I used a function, as shown here, to calculate RE. Accordingly, RE is equal to 0.857 for 22 degree C. This means that 22 degree C hardly fulfills the requirement “cold temperature.”
The total information content of a solution is a point ( CE , RE ) in the plot RE vs CE . Since low entropy is good, solutions having both RE = 0 and CE = 0 are good solutions. Here you can see the position of three selected temperature.
Sometimes a solutions might have many points ( CE , RE ). In this case, i can construct a boundary and see how far it is from the origin. This farness is given by a parameter called coherency measure denoted by lemda. I can minimize lemda to find optimal solution. The case shown in this slide, solution A is better than solution B because coherency measure of A is small than that of B.
Let me introduce you the customer needs assessment.
Recall the slide No 15 ,where i emphasized is customer needs. When the internal customers interact with the external customers for the sake of key solution determination, the following issues arise: What is the appropriate customer need model? How to deal with the unknown customer needs? How to classify the key solutions based on customer responses? Now, I will show you , how to answer the above-mentioned issues.
Let me introduce you the roadmap of customer needs assessment.
I went to Dhaka, the capital of Bangladesh to collect customer opinion on some features of passenger vehicles. 100 respondents answer the questions on 38 features.
Let me introduce you the some sample feature out of 38 features of small passenger vehicles.
I prepared questionnaires according to Kano model. First I can ask a question when a feature/function is present. The options are like, must-be, neutral, live-with and dislike. The respondent chooses one. Then I can ask another question when the feature/function is not present. The options are: same. Again the respondent chooses one.
Based on the answers, a feature is classified into Attractive (A), Indifferent (I), Must-be (M), One-dimensional(O), Questionable (Q), or Reverse (R).
The meaning of Attractive (A), Indifferent (I), Must-be (M), One-dimensional (O) and Reverse (R ) are shown here.
Recall slide -27, I have used discrete event simulation to know the unknown answers of external customers. Here, Z is the simulated event which can be either Like, or Must-be, or Neutral, or Live-with, or Dislike. To simulate Z, i can use five rules. The rules use the cumulative probabilities of the possible states of Z. The cumulative probability is calculated from the Probability. The Probability is calculated from the Expected Value of Linguistic Likelihood. The Expected Value of the linguistic likelihood is calculated from the linguistic likelihood of relative frequency. The linguistic counterpart of relative frequency is the linguistic likelihood corresponding to the maximum membership value. Random is a uniformly distributed variable in the interval [0,1].
Here I show you are example of customer answers. According to customer answer the feature is an indifferent feature.
Why it is indifferent? The reasons the answers are mostly Neutral.
If I delete the neutral answer , then I see the Kano– evaluation becomes Reverse.
Reverse is not a realistic conclusion for Sedan feature. This means that I couldnot use simple frequency based analysis.
Therefore, I revisited the classification of feature. I can define a feature , that is, a must be included feature if it is either O, or M. I can define a feature , that is, a should be included feature if it is A. I can define a feature , that is, a could be included feature if it is either not I, or not R. I can define a feature , that is, a unreliable feature if it is Q.
Based on that i classified the customer answers , I found that Sedan has low information content , when it is considered should –be included feature.
I found that SUV has low information content , when it is considered must–be included feature.
I found that Van has low information content , when it is considered could –be included feature.
This slide shows a preferential chart of customer choices regarding cars selection. To deal with the intrinsic complexity of customer needs analysis, logical aggregation of customer opinions is a better choice compared to that of frequency based analysis. This faculty of thought is demonstrated to be true by logically aggregating the field data of customer needs collected from Bangladesh on small passenger vehicles. The multi-valued logic plays an important role in the logical computation. For the sake of a better understanding, Kano-model-based customer answers are considered.
Let me introduce you the sustainability issue.
Recall the slide No 14. where I emphasized is sustainability. Sustainability has earned a great deal of importance because of growing environmental concerns. I will show you now h ow to deal with sustainability in key solution determination process .
I considered the sustainability of grinding wheel . Sustainability of the grinding wheel depends on the environmental effect of primary production of hard materials.
From this point of view, I can try identify the eco-attributes of hard materials. Here you can see the Eco-attributes (water uses Vs CO 2 foot print) of primary production of hard materials ( i.e. technical ceramics).
I studied five materials namely, Alumina, Zirconia, Silicon carbide, Boron carbide and Boron Nitride. Here you can see their eco-attributes means CO 2 footprint, water usage, NO x , SO X ,............ . Here information is imprecise and is given by a numerical range. So, I can use the concept of range compliances as I explained before. (Four eco-attributes called CO 2 footprint, NO X emission, SO X emission and Water usage of five classes of hard materials based on Alumina, Zirconia, Silicon Carbide, Boron Nitride, and Boron Carbides are studied. In all cases, numerical ranges give the eco-attributes, not by a sharp data points.)
Let me introduce you the universe of discourse for two cases for sensitivity evaluations of the materials. To deal with the imprecision associated with the eco-attributes, an entity called range compliance is used. The compliance of an eco-attribute given by a numerical range is determined by using a set of five linguistic classes labeled very low, low, moderate, high, and very high.
Let me introduce you the evaluation of Alumina ; which is close to the good zone of environmentally friendly materials according to the slide No 25.
Let me introduce you the evaluation of Zirconia; which is some cases close to the good zone of environmentally friendly materials some cases not according to the slide No 25.;
Let me introduce you the evaluation of Silicon carbide; which is some cases close to the good zone of environmentally friendly materials. material some cases not. according to the slide No 25.;
Let me introduce you the evaluation of Boron Nitride; which is relatively far from the good zone of environmentally friendly materials.
Let me introduce you the evaluation of Boron Nitride ; which is mostly far from the good zone of environmentally friendly materials.
Let me introduce you the relative positions of environmentally friendly and less friendly materials. It is found that Alumina based hard materials has low environmental impact followed by that of Zirconia and Silicon Nitride base materials. Boron Carbide has the highest environmental impact. The environmental impact of Boron Nitride based materials remains between that of Silicon Carbide and Boron Carbide based materials.
Let me introduce you the creativity assessment for the key solutions of the conceptual phase of product development.
Recall the slide No 13, where, I emphasized that How to differentiate a creative-key-solution from a non-creative-key-solution is one of the decisionmaking challenges for internal customers. Now, I will show you some results on this issue.
Creativity has many definition, one of the recent definition is given in C-K theory. C means concept, K means Knowledge. When creative concept means a concept which is undecided in the beginning.
Now, I will show you the creativity status of two solutions C1 and C2 . C1 means an existing engine, C2 means a creative engine. Both are for mars exploration. I define 4 propositions for C1 and 4 propositions for C2.
This slide shows the information content of existing engine for mars exploration.
Let me introduce you 4 propositions for C2.
This slide shows the information content of creative engine for mars exploration.
Thus, conclusion is that ........................................ Creative concept means a concept which is undecided in the beginning of the conceptual phase of product development. Information content of design from the sense of epistemic uncertainty should be maximized to remain creative.