Loading

Module 1: Physical & Operational Architecture Development

Notes
Study Reminders
Support
Text Version

Methods of Allocation of Function

Set your study reminders

We will email you at these times to remind you to study.
  • Monday

    -

    7am

    +

    Tuesday

    -

    7am

    +

    Wednesday

    -

    7am

    +

    Thursday

    -

    7am

    +

    Friday

    -

    7am

    +

    Saturday

    -

    7am

    +

    Sunday

    -

    7am

    +

In the last class we discussed about the operational architecturebasically looking at the allocation of a function to components and how do we actually do theallocation architecture, what are the principle of allocation architecture or how do weallocate, what are the basic principle to be used. And then we briefly mentioned that thisallocation architecture helps us to identify the tradeoff decisions as well as look at the inputand output requirements, which are coming within the system.In this class we will look at allocation architecture in bit more detail and try to identify howdo we get the derived requirements input output requirements coming because of theallocation and how do we do the tradeoff analysis and how do we get tradeoff decisions basedon the allocation architecture.(Refer Slide Time: 01:12)
So, to going to the details in the last class we mention that 1 of the major task in allocationarchitecture is basically to look at the non input output requirements. That is the input output
requirements; we know that the from looking at the system, as a whole we have some inputswe have some outputs and based on that we can identify the input output requirements.But, when we develop the allocation architecture, when we do it from the physicalarchitecture convert that into from the using functional architecture and physical architecture,when we convert that into allocation architecture there will be some other requirementscropping up because of this allocation; and they are basically the internal input outputrequirements.Basically we look at what are the inputs and the outputs internal to the system and we need toallocate the components to for these requirements the other one is the system widerequirements there are system wide requirements identified at the design level itself, but thenhow we actually allocate these requirement system wide requirements of a cost reliability,availability of the system. So, these things need to be allocated to the component. So, how dowe actually do the allocation of system wide requirements or identifying the system, widerequirements, coming out of the allocation architecture and then how do we allocate them.The other one is how do we get the tradeoff requirements and then how do we get thequalification requirements of the system.So, all these things need to be analyzed using the allocation architecture. So, we go 1 by 1 welook at the internal input output requirements.(Refer Slide Time: 02:48)
And how do we actually trace these requirements and then get the allocation architecture. So,the internal input output requirements.(Refer Slide Time: 02:54)
Basically we would look at the example given for the elevator system we can see that whenyou allocate the component or the system you can see these are the main function A 1, A 2, A3, and A 4 in terms of IDEF0 diagram.(Refer Slide Time: 03:11)
So, this is A1 this A 2.
(Refer Slide Time: 03:13)
This is A 3 and there you have A4 function.(Refer Slide Time: 03:15)
So, if from these function there will be some other requirements coming up, that is theinternal input output requirements we look at these you can see digitized passenger requestsis 1 of the output from this function, which is going as an input to the control elevator cars.So, this is an internal output from this function as I said internal input to these function andthis is the other one assignment of elevator cars is another 1 and then you have anotherrequirement of elevator position and direction. So, this is actually coming as an output from
the move passenger between floor function and similarly you will have other one morerequirement like the sensed malfunction and for kind malfunction are that need to be sensedwithin the system.So, if you like this you have few input and output requirement which is internal to the system.So, for every system whenever you develop and allocate the architecture for thesecomponents, we need to find out which is the components which actually satisfy thisrequirement these input output requirement or to which components these requirement needto be traced. So, 1 thing is basically do trace the requirement and then other one is to allocatethese input output requirements.So, here we can see there are 4 input output requirements.(Refer Slide Time: 04:34)
Like here it is digitized passenger requests assignment for elevator cars and elevator positionand direction. So, these are the requirements identified and the requirement is the functionwhich actually needs to be provided the requirement can be written as, the elevator systemshall produce digitized passenger requests or the elevator system shall consume digitizedpassenger requests. So, these are the requirements which can be written using this functiondecomposition and these are the internal input output requirements, which cannot beidentified from the system level. This can be identified only when we have a functionaldecomposition then try to allocate these requirements.
These input output need to be traced to corresponding items and the function responsible forconsuming and creating the item respectively, it is not only identifying the requirements weneed to find out to which component or to which system these requirements can be traced.So, that is the tracing of the requirements. So, whatever the requirements we identify. So, thisis just a sample only these requirements there will be many such requirements input outputrequirements, we need to identify the components or we need to trace these requirements tothe corresponding items and the functions or responsible for consuming and creating this kindof input and outputs.(Refer Slide Time: 05:56)
So, here you can see that the elevator system shall produce digitized passenger requests thatis 1 of the requirements, this can be traced to the function accept passenger request andprovide feedback. So, this is the traceability of the requirements, similarly the elevator systemshall consume digitized passenger requests which can be traced to the function controlelevator cars.Similarly we can actually trace other requirements also to the corresponding functions. So,not only the functions what we need to do is to look at now which function or whichcomponent is providing this function. Then the requirement can be traced to that particularcomponent also. Therefore, we can have a traceability of the input output requirements whichis internal to the system, as well as we can trace these requirements to the correspondingfunctions and components.
So, the allocation architecture basically we will look into the details of input outputrequirements coming from within the system and then how do we trace these requirements tothe corresponding functions and components. So, these are the just 2 examples of how do wewrite down the requirements, input output requirement and then trace this requirement to thecorresponding functions and components.(Refer Slide Time: 07:12)
Other 1 is tracing the system wide requirements and deriving sub system wide requirements.So, the first 1 what we discussed was the input output requirements, the next version isbasically the system wide requirements, as I told you the system id requirements can be a costit can be a reliability of the system or it can be availability or durability, which is identified atthe stage of overall system design then we need to actually trace these requirements, as wellas allocate these requirements to components basically when we said cost cannot be given toyes particular component. So, this has to be allocated to sub system. So, how do we actuallytrace the requirement of this particular system wide requirement and then allocate themamong the components or if you take the reliability of the system depends on the availabilityof different components of the system.So, how do we actually do an allocation of this reliability amongst these components and thenensure that the total reliability of the system is maintained. So, the task of allocation ofreliability is basically to look at the various components, which actually contribute to theparticular requirement. And then allocate the values accordingly to the components, then the
allocation architecture by looking at the architecture we should be able to allocate thesevalues and then trace this value to the corresponding functions and components.So, that is the system wide requirement on cost reliability, availability, durability etceteraneed to be allocated among the components of the system and how do we do this is anexample whether the system shall cost rupees 10000 or less to use per month during it isoperation or the system shall employ ABC technology. So, these are the typical system widerequirement, like this you can have the system should have a reliability of 0.96 with a designgoal of 0.99 that can be another system wide requirement.So, this is the system wide requirement, when you say the system shall cost 10000 or less touse per month during it is operation, we need to find out what are the things actually willcontribute to the cost that is operating cost and then how do we allocate this operating costamong the components. For example, if you take the elevator system itself the operating cost,may include the power charges, it may include the control cost, it may include the inputoutput requirements.So, all those will be contributing to the total operating cost and this we need to allocateamongst the other component. So, that is the requirement over here similarly the system shalluse a particular technology. So, this actually we need to see if there is a particular technology,then how do we actually identify the component is actually require the same technology to beused or a technology which is compatible to the suggested technology. So, that way we needto allocate this system wide requirement.There are different ways of doing this. So, the question here is how do we allocate thesystems wide requirements to component or subsystems, that is a one of the important tasksin allocation architecture, we need to allocate these system wide requirements tocorresponding components.
(Refer Slide Time: 10:20)
So, how do we do this? There are a different ways of doing this, one is known as apportionedmethods, the other one is known as equivalence method the third one is the a synthesismethods. So, these are all by Grady to propose in 1993.So, we can use any one of these methods to allocate the system wide requirements to it iscomponents. If you look at the apportionment methods, it is basically providing the actualsystem wide requirements in a proportional manner to different components. So, the mostappropriate for this kind of methods is the cost requirements, reliability, availability,durability etcetera. So, this can actually be apportioned among the components. If there are 5components which actually contribute a particular requirement, then we find out what shouldbe the contribution from each component and accordingly we divide the requirements, thesystem wide requirement and then distribute them amongst the member. So, that is allocatethe requirement to the members who actually contribute to that particular requirement.So, here the system level requirement is divided or apportioned out to the systemscomponents not necessarily in equal increments, but keeping a margin of 5 to 10 percent asrisk mitigation strategy. So, what we do here is, we look at all the components and then aportion this the main requirement to amongst these components, but I will we do not do it inan equal manner, we will look at the component and then decide what should be the otherproportion to which it should be divided.
So, it need not been equal increments, but at the same time we do not give the completerequirement to a all the components, what we do is we keep a 5 to 10 percent margin, overhere and then this margin is used for the risk mitigation strategy. So, if you want to avoid therisk at the later stage keep a 5 to 10 percent margin and this margin will be used at later stageto see whether (Refer Time: 12:21) particular risk is coming from any other component. So,that can actually be compensated using this method that is the apportioned method.One example you could be the cost of operation of the elevator, if you say that it 10,000rupees per month for operating the system, that is the maximum allowed then we can do 10percent as a margin. So, maybe for 9000 we can identify, what should be the maximum costcoming from the components. So, 9000 rupees can be divided amongst these 4 componentsdepending on the importance of each one probably the for the maintenance can actually have1000 rupees, in the co`ntrol elevator cars could be for 7000 rupees and then the remaining1000 rupees can be divided amongst other members.So, the total 9000 operating cost can be apportion to amongst the members, and 1000 can bekept as a risk mitigation strategy. So, that is the simple way of allocating the system widerequirements in a system. So, here as you can see as I mentioned it is not in equal incrementthat is 1 of the points you need to note down and again we will need to keep a margin of 5 to10 percent for risk mitigation.(Refer Slide Time: 13:27)
This is another example here the system wide reliability of elevator is 0.9 with the designgoal of 0.99.So, what we are saying here is that, there is a system wide requirement of reliability which isto be maintained at 0.9 and with the design goal of 0.99. So, this is the minimumrequirements and the goal is to get a very high reliability of 0.99. Now we have differentcomponents in the system and we need to find out what could be the reliability of thesecomponents in order to get the; I total reliability of 0.9.If you take the functions in the elevators in A 1, A 2, A 3 and A 4. So, these are the 4 mainfunctions in the elevator, and we need to have the reliability of these 4 combined togethergetting a 0.9, that is the requirement. But we do not know how much we should give thevalue for this and for this, this and this. So, these are the 4 components and we can do it in athe apportionment method we will try to divide these reliability amongst the members,depending on the our knowledge about the system and how it is going to affect the holdoverall system, accordingly we will try to provide the reliability values for this.So, in this case apportioned methods what we do is.(Refer Slide Time: 14:45)
We write down the requirements reliability of components like a passenger interface can havea reliability of 0.96 with the 0.996 as the design goal. Similarly elevator controller can have areliability of 0.995 and with the design goal of 0.995 and elevator compartment can have a
reliability of 0.96 with a design goal of 0.96 and elevator maintenance can have a 0.99 andthe 0.99 as the design goal. If you multiply all these reliability we will be getting it as 0.91;that means, there is a risk mitigation value of 0.01 for later use.So, the total a reliability of components will be 0.91. So, you have a 0.01 or a clearance or thetolerance for risk mitigation. So, this is the way how we do it in apportioned methods andthen how do we get this is again it comes from various factors, if you know the particularcomponent you are using, you have to use the reliability value provided by the manufacturerof this particular component or if you know that the corresponding system or there aremultiple systems coming in this one. So, we need to look at those systems and it is reliabilityvalues and accordingly it is fix a value for this one. Again this is to be done by the designteam. So, there is a little bit of subjectivity in this one, but still it is a simple method. So, thatis why it is used for the system requirements and allocation of the system wide requirementsamongst the components. So, that is the way how the apportionment method works.So, here once you have this then we can actually identify the derived reliability requirement,the elevator component passenger interface which is this passenger interface shall have areliability of 0.96 or greater, the design goal is 0.996. So, once you have this one, then youwill write down the derived requirement which is coming from the apportioned methods foreach component we can write the derived the requirement as the system or the elevatorcomponents, shall have a reliability of 0.96 or greater the design goal is 0.996.So, this is the way how we give the reliability, how we use the apportioned method to dividethe reliability or allocate their system wide requirements to the components.
(Refer Slide Time: 17:07)
Another method is known as equivalence methods, and this method as the name suggests it isbasically divided it equally amongst the components. So, the component requirements sameas system requirements, there is no difference, but all that system requirement is there thesame requirement will be given for the components also. The system constraints areappropriate for equivalents method. So, constraints in terms of when you design the systemwhere will be the system requirements as well as the system needs as well as the constraints.So, what we do here is to use the system constraints basically to we use the equivalentsmethod to allocate constraints to the components. Basically if you look at requirements likethe system shall have a olive green in color. So, system shall be olive green color if that is arequirement, then we need to make sure that all the components are of the same color. So, thisis known as a simple equivalence method of allocation.The third one is a synthesis methods, which is used when the system level requirement iscomprised of complex contribution from the components, causing the componentrequirements that are flowed down from the system to be based upon some analytical model.So, this is more of an analytical method of allocation. So, here difficult to identify thecomponent requirements because they are related to different components and eachcomponent is having different way of representing it is performance.So, in this case then we need to have some analytical model to do this. The system levelrequirement has significantly different units than the derived component requirements has.
The system level requirement and the component level requirement may be having differentunits. So, in that case we will not be able to directly allocate these requirements to thecomponents and therefore, we go for an analytical method. For example, if there is arequirement of system performance time. If you say that the system should provide an outputwithin 20 seconds that is the requirement, but then the processor is one of the componentwhich actually provide this, but the processor speed is different unit and there may be someother motors or something which actually provide the output and it is output will be in adifferent unit.So, in these cases we need to ensure that the requirement at the system level is given in adifferent unit, but the components are in different units. So, we need to have some kind of amathematical modelling or an analytical method to allocate these system wide requirementsto component. So, there we go for the synthesis method. There are different synthesismethods, we will see the methods at later stage, but this is one of the important methods bywhich we allocate the system wide requirements.(Refer Slide Time: 19:47)
The example is the requirement relating to the average time between the passenger making arequest, and being delivered to the requested floor needs synthesized between all the 4components which provide the service.Suppose if you have a requirement like this, average time between the passenger making arequest and being delivered to the requested floor. So, this needs to be synthesis between the
all the 4 components and each component of this system will be having different way ofproducing an output. Because, it takes a function of the input output interface, the passengerinterface there will be a sometime delay in passenger interface and there will be a time delayin the motor output and the acceleration of the elevator car and other elements also.So, we need to see what kind of a relationship exists between the total time for service as wellas other component performance parameters and once we have this mathematicalrelationship, then we try to identify or try to have a relationship and then see how to allocatethe main requirement to this component. So, that is the method of Synthesis methods.(Refer Slide Time: 20:53)
The next one is basically the trade-off requirements and subsystem trade of requirements.Basically we look at the requirements like cost, schedule, performance etcetera there will besome trade-off in the system because of this. Mainly if you want to reduce the cost oroptimize the performance, we need to make some kind of a tradeoff between these parametersthe cost the scheduled the performance etcetera.So, how do we trace this trade-off requirements and then how do we derive the sub systemtrade-off requirement, because the system will be having some trade-off and sub system willbe having at different trade-off. Because there are many other features which actually comesinto picture here, the objectives hierarchy what is the system objectives and how the variousobjectives are given the particular value in the hierarchy, the traders requirement basicallydepends on those values also. So, we need to look at the hierarchy of the objectives and based
on the objectives hierarchy, we need to decide what kind of tradeoffs to be provided in thesystem.(Refer Slide Time: 21:57)
For example if you take the again the elevator case study, you can see that the objectiveshierarchy is given here the operational objectives like operational performance and monthlyoperating cost. So, these objectives are given here you can see that the operationalperformance objective weight is 0.9, but the monthly operating cost weightage is 0.1. So,here the more weightage is given for operational performance. So, whenever we have atradeoff, we need to make sure that the performance objective values are met then the costobjectives.So, we can always have a trade-off with cost to get a better operational performance. Andwhen we look at the operational performance, we will see that a there are many other subobjectives coming from the subsystem like time in systems. So, now, to get the operationalperformance objective 0.9, we have a timing system objective of 0.35, then we have a rightquality objective over here 0.3, and then availability objective of 0.35. So, these objectives tobe met when we have the trade-off; so whenever we have a tradeoff analysis, we need to lookat the objectives hierarchy and accordingly we need to find out for how we actually allocatethese objectives and then decide on the tradeoff of the system wide requirements.
Again different ways of doing this; so we can actually as I told you the cost scheduleperformance all these can be actually be traded off for various improvement in eitherperformance or the cost or schedule.(Refer Slide Time: 23:37)
So, this can actually be done by different methods, the decision analysis for design trade-offthe standard methods, where we have a multi attribute value analysis which is known as MVAor we can use an analytical hierarchy process. So, any one of these methods can be used forthe tradeoff analysis. The multi attribute a value analysis is one of the most common methodused to find out how do we actually allocate the requirements and the based on them valuefunctions of the system.
(Refer Slide Time: 24:07)
So, here this is a quantitative method for aggregating a stakeholders preference overconflicting objectives to find the alternative with the highest value when all objectives areconsidered. So, this is analytical method use in order to get a an alternative with the highestvalue, when all objectives are considered. So, you have multiple objectives and they are littlebit conflicting. So, in this situation how do we actually get the alternative, the best alternativewith the highest value and all objectives are considered. So, that is the multiattribute valueanalysis.So, the various steps involved in multiattribute value analysis are basically we define a valuescale for each objective, that is what should be the scale on to which the objectives tomaintain. So, what is the minimum value to be and what is the maximum you can have, andthen quantify the relative value of improving from lower to higher level in the form of valuecurves. Now, we have a minimum value and a maximum value, and what kind of a variationwe are expecting from this minimum to maximum or it is a linear variation or it is anexponential variation, understand increasing or decreasing variation.So, what kind of a variation we are expecting that is to be given in terms of value functions?And then we need to normalize this value function, because the values are different scales.So, we need to have an organized value function and once we have the normalized valuefunction, we can actually have a relationship which will actually optimize the total valuefunction and that value function can be used for optimizing the objectives.
So, normally exponential functions are used to approximate normalized value functions. Thisvalue function as I mentioned these value curves, these are the value functions. So, normallyexponential functions are used to approximate the normalized value functions and once wehave these exponential functions by controlling some of the parameters, we would get variouscurves depending on the requirements.(Refer Slide Time: 26:05)
This is the exponential function normally used and the normalization of function is doneusing this method.(Refer Slide Time: 26:10)
So, we get the value functionvi ( xi)=1v`i( ximax )−v`i( xi0)[ v`i ( xi)−v`i( xi0)]
vi( xi) prime based on the exponential curve, and then we normalize it along the differentvalues that is the maximum value and the minimum value, and the value of ( xi) and xi0.So, this actually will give you a normalization of the function. So, this normalized functionwill be used for getting the total value function that is
vi ( xi)=1−e−α ( xi−ximin)1−e−α ( ximax−ximin)
V (x )=∑i=1nwivi ( xi)
vi ( xi) So, if you have many value functions any value functions. So, we will take the totalvalue function as
V (x )=∑i=1nwivi ( xi)
So, the vi( xi) is the value function, wi
is the weight of the value function.
We will take a simple example and tell go through these steps and I will tell you how to dothis. So, we will go to the board and then explain how do we actually use the value functionsand value curves to get the total value function, when we have multiple value functions ormultiple objectives in a system. So, as I mentioned the first task is basically do define thevalues.
(Refer Slide Time: 27:16)
So, define the values basically for each objective. So, we can say that the timing system isbetween. So, the minimum value is 3 seconds, and the maximum is 10 seconds. So, we canactually define this kind of minimum to maximum values for any function. So, it can be 3 to10 seconds or you can say the cost which should vary from 9000 minimum cost is 9000 andmaximum we can go for is up to 10,000 rupees.