Rationalization of model development Illustration
Modeling of organization contribution on support performances
MARC CYRILLE FABIEN CHAUCHAT
KTH ROYAL INSTITUTE OF TECHNOLOGY SCHOOL OF ENGINEERING SCIENCES
Rationalization of model development Illustration
Modeling of organization contribution on support performances
MARC CYRILLE FABIEN CHAUCHAT
Degree Projects in Systems Engineering (30 ECTS credits) Degree Programme in Vehicle Engineering (300 credits) KTH Royal Institute of Technology year 2019
Supervisor at Dassault Aviation, France: … Supervisor at KTH: Per Enqvist
Examiner at KTH: Per Enqvist
TRITA-SCI-GRU 2019:329 MAT-E 2019:81
Royal Institute of Technology School of Engineering Sciences KTH SCI
SE-100 44 Stockholm, Sweden URL: www.kth.se/sci
Report
Rationalization of model development
Illustration: Modeling of organization contribution on support performances
Rationalisering av modellutveckling
Illustration: Modellering av organisationens bidrag på supportprestationer
DOMAINE
Provisoire
Direction Générale du Soutien Militaire
Edition Date Indice Rédacteur Visa Rédacteur Visa Approbateur
Origine 26/03/2019 Marc CHAUCHAT
Dernière mise à jour 06/09/2019 Marc CHAUCHAT
Chauchat Marc Gilles Debache
page : 1 / 68
Ce document est la propriété intellectuelle de DASSAULT AVIATION. Il ne peut être utilisé, reproduit, modifié ou communiqué sans son autorisation. DASSAULT AVIATION Proprietary Data.
DGSM 254143 06/09/2019
Classement : Document3
TABLE OF CONTENT 2 Provisoire
TABLE OF CONTENT
1. PRELIMINARIES ... 4
1.1 Document history ... 4
1.2 List of modified pages ... 4
2. ABSTRACT ... 5
3. ABSTRAKT ... 6
4. ACKNOWLEDGMENT ... 7
5. INTRODUCTION ... 8
6. ENGINEERING ... 9
6.1 Engineering and ingenuity ... 9
6.2 Engineering and optimization... 9
6.3 Engineers and organization ... 11
6.4 From optimization to innovation ... 12
6.5 Summary ... 13
7. MODEL DEVELOPMENT ... 14
7.1 Development cycle of a model ... 14
7.2 Models and data structure ... 14
7.3 Deductive and inductive methods ... 18
7.4 Expertise phase ... 19
7.5 Sensitivity analyses ... 22
7.6 Incremental development ... 23
7.7 Industry specificities ... 24
7.8 Summary ... 25
8. INTERNSHIP ... 26
8.1 Introduction ... 26
8.2 Characterization of a model ... 30
8.3 Impact of the organization on support performances ... 34
8.4 Cameo Enterprise Architecture model ... 44
8.5 Excel model ... 50
8.6 Results ... 51
9. INTERNSHIP APPRAISAL ... 53
9.1 Company contribution ... 53
9.2 School contribution ... 53
9.3 Possible continuations ... 54
9.4 Personal benefits ... 55
TABLE OF CONTENT 3 Provisoire
10. CONCLUSION ... 56
APPENDIX A. GLOSSARY ... 57
APPENDIX B. BIBLIOGRAPHY ... 58
B.1 Availability ... 58
B.2 Cost ... 58
B.3 Engineering & methodology ... 58
B.4 Metrics ... 58
B.5 Organization ... 58
APPENDIX C. DASSAULT AVIATION ... 59
APPENDIX D. CHARACTERIZATION OF THE IMPACT OF AN ORGANIZATION ON SUPPORT PERFORMANCES ... 63
D.1 Needs ... 63
D.2 Development and governing team of the model ... 65
D.3 Data typology ... 65
APPENDIX E. EXCEL MODEL ... 67
Provisoire 1. PRELIMINARIES 4
1. P RELIMINARIES
1.1 Document history
Date Indice Rédacteur Objet de la mise à jour 26/03/2019 Marc CHAUCHAT Création du document
1.2 List of modified pages
Toutes les pages de cette édition sont au dernier indice du document Sans objet pour cette édition
Provisoire 2. ABSTRACT 5
2. A BSTRACT
This report is resuming the internship I did during Q2 and Q3 2019 at Dassault Aviation on the subject of standardizing the modeling process. In particular, this process needed rationalization.
This report is also containing a presentation of my understanding of engineers’ job and engineering work within this particular industry: Dassault Aviation being one of the top main French aircraft constructors, both with business jets and military aircrafts.
This work is based on a real case that will roughly be described here too: during the internship, a model aiming at representing the impact of an organization on support performances of an aircraft has been developed. By changing parameters values to evaluate the impact of an organization for different configurations, it has been computed that this impact was negligible for short (<1h) and long (>4h) turn-around duration while being maximum for a turn-around cycle lasting between two to three hours.
When developing this model, I could myself experiment different technics aiming at optimizing the development of models: incremental design, organizing specific models through a generic data structure, doing a sensitivity analysis on parameters of a model to allocate resources for an efficient development.
Even though this work was made for the support of military aircrafts, it can be extended to a civilian environment, or for drones.
Provisoire 3. ABSTRAKT 6
3. A BSTRAKT
Denna rapport beskriver den praktik som jag gjorde under Q2 och Q3 2019 på Dassault Aviation och hanterar om standardisering av modelleringsprocessen. I synnerhet kräver denna process rationalisering.
Den här rapporten innehåller också en presentation av min förståelse för ingenjörers jobb och ingenjörsarbete inom just den här branschen: Dassault Aviation är en av de främsta franska flygplanskonstruktörerna, både med affärsflygplan och militära flygplan.
Detta arbete är baserat på ett verkligt fall som grovt kommer att beskrivas här också: under praktikperioden har en modell som syftar till att representera en organisations påverkan på supportprestanda utvecklats. Genom att ändra parametervärden för att utvärdera effekterna av en organisation för olika konfigurationer, har det beräknats att denna påverkan var försumbar under korta (<1 timme) och långa (> 4 timmar) vändningstider, medan den var maximal för en vändningscykel som varar mellan två till tre timmar.
När jag utvecklade denna modell, kunde jag själv experimentera med olika tekniker som syftar till att optimera utvecklingen av modeller: inkrementell design, organisera specifika modeller genom en generisk datastruktur, göra en känslighetsanalys av parametrar för en modell för att fördela resurser för en effektiv utveckling.
Även om detta arbete gjordes för att stödja militära flygplan, kan det utvidgas till en civil miljö eller för drönare.
Provisoire 4. ACKNOWLEDGMENT 7
4. A CKNOWLEDGMENT
I would like to thank, within Dassault Aviation:
My tutor for his support during the internship, his ideas and discussions as well as his happiness that made working together an interesting moment and a great experience.
All co-workers that helped during the internship via technical points and every co-worker for it kindness and discussions that made it a pleasant and fulfilling experience.
All members of human resources that made possible this internship and ensured a smooth development of it.
I would also like to thank, within KTH:
Per Enqvist, for his support and coordination with KTH
Provisoire 5. INTRODUCTION 8
5. I NTRODUCTION
Over the past few years, industrials have been developing more and more models to be able to represent reality in a detailed and accurate way. They have changed from conventional to digital means to support their models and ideas. This transition is due both to the complexification of systems and the globalization of the economy and the market. In fact, systems are harder to understand due to their complexity and industrials are challenged by more competitors and need to enhance their offer towards customers to stay leaders in their domain. This competition along with limited resources forces industrials to be more and more efficient.
As a consequence, the rationalization of model development is a key point for an efficient use of resources when developing digital solutions. Several solutions to implement this rationalization have been observed during the internship: from incremental development which develop a model deeper and deeper without impacting other parts of the model, to sensitivity analyses, that are part of iterative design, where one will allocate resources to develop the most valuable parts of a model. An architecture based on a global data structure also helps organizing interactions with specific models therefore increasing the efficiency.
Even though engineering can be described in many ways, a simple approach would be to consider it as a rational answer to a need. One should know that the core activity of engineering is ingenuity even if engineers have been used mainly for optimization purposes since the steam revolution.
Hence, this document will first go through engineering history and the understanding of engineers’ job I developed through observations and discussions within Dassault Aviation.
Following this, my understanding of methods use in the aerospace industry for the development of models will be developed. These methods will be illustrated by a concrete example of the model I developed during the internship about the impact of an on-site organization on the support performances of an aircraft.
The model developed intends to include all main factors impacting the operational availability of an aircraft for a pre-design study. Both technical (aircraft design, support, logistics) as well as non-technical (organization, administration) factors were looked at.
Provisoire 6. ENGINEERING 9
6. E NGINEERING
This section will defend that the core activity of engineering is ingenuity and follow the evolution of engineers over the years to realize that their activity has been moved towards optimization and organizations purposes. One will see that this evolution now turns into a cycle where engineers go back to their innovative function thanks to AI1, big data and knowledge management.
6.1 Engineering and ingenuity
Ingenuity is the essence of engineers. It lies in the capacity to invent solutions to a problem using external means, ingenuity is not about using inherent functions. Developing new solutions can be achieved by adding some characteristics to an existing element or by using it in a new original way. For instance, one can use a rock as a hammer or one can shape a rock to obtain a knife and combine it with a wood stick to obtain a spear. Most of the time, it lies in the possibility to create a new approach to answer a need. By doing this, an engineer opens new perspectives for some users (the moon walk) or an entire society (with the invention of the Internet network).
First one should know that ingenuity is not specific to humans as some animals also shown they are able to add value to some objects or use them in a new original way. For instance an ape can use a rock as a hammer.
One can see here that the essence of engineering and therefore engineers’ job is to formally develop the native human capacity at creating responses and solutions to solve a problem.
6.2 Engineering and optimization
6.2.1 Introduction
Engineers have been developing solutions and systems for thousands of years. One can see testimonies of engineering solutions in histories written at that time such as the Trojan horse developed by Ulysses2.
Over the next centuries, engineers have developed systems that even though they seem simple nowadays were complex at the time due to limited knowledge and resources or technics. For instance a wheel might seem simple today but has been a huge breakthrough
1 Artificial Intelligence
2 During the Trojan War, no matter how they organized their armies, Greeks could not break the Trojan walls.
After years of war, Ulysses had the idea of making the Trojan horse to go through these defenses and take the city. One can see Ulysses as one of the first engineers as he used his ingenuity to deceived Trojans turning a strategic artefact to end the war into a harmless offer to Gods.
Provisoire 6. ENGINEERING 10 at the time allowing people to travel easily. These breakthroughs changed people’s life and shaped the world as we know it.
Some inventions were major steps and radically changed the world. For instance the steam revolution has been a decisive turn in history, granting people almost unlimited energy compared to what they had and allowing them to travel even further. With this ability to travel far away, people were able to share their knowledge easily. One will now see how this share of knowledge changed the life of engineers.
6.2.2 Engineering since the steam revolution
Before the steam revolution, knowledge was most of the time gained through experience and if learned from someone else often limited by geography and movement because people did not travel that much. Knowledge was nevertheless varied, from architecture and biology to craftsmanship, but was not shared and discussed between many other wise persons.
With the steam revolution, people were able to move faster and further, hence knowledge was harvested and shared massively. As a consequence, a lot of innovations were available for people. Because people asked for better versions of such systems, engineers have been pushed towards optimization and it has therefore reduced the time allocated for research of breakthrough innovations.
Due to the steam revolution, “energy” was conceptualized. The concept of energy allowed decoupling the coal and the locomotive yields. One could then speak about the energetic yield of coal, as an indicator of its quality, and the yield of the machine, which represent how well it extracts and uses this energy. From this moment, engineers’ job has been focused on optimization, mainly a yield optimization. That is to say engineers try to reduce all leaks in order to orient flows towards the same direction. This massive access to energy has led to a lot of innovations, for instance mechanizing a lot of activities. Nevertheless the breakthrough innovation was the steam machine.
Even though systems were complex to imagine and build in the past, the steam revolution and an easy access to knowledge and technics lead to more and more complex systems.
Nowadays, even with a simple access to an immense source of knowledge accumulated through history, a single person cannot understand complex systems globally while understanding every detail of them. For instance aircrafts and their surrounding environment constitute a very wide and complex area. This can be explained as one need to learn a lot and needs to specialize oneself3. This specialization helps to understand very complex phenomenon. The limit is that one can only have a general idea of other specific areas he did not study.
Over time, other breakthrough innovations have emerged: a secondary energetic revolution with oil and electricity and a numerical one more recently with computers. Nowadays, people are limited by their ability to analyze data. Nevertheless one can see that with AI, big data and knowledge management, people are able to analyze gigantic amount of data to find an optimization solution faster than before. These innovations are thus reducing the
3 Jean Baudet, Histoire des techniques
Provisoire 6. ENGINEERING 11 optimization load for engineers allowing allocating more time for innovation. Nevertheless, optimizing systems has not been the only activity of engineers. They had to interact with organizations which emerged due to the complexity of systems.
Over the years, engineers have developed tools and they are still spending time working on tuning systems in order to achieve the best energy ratio they can. That means engineers have studied ways to achieve the same result using as small resources as possible or achieve a better result with same given resources. This is the philosophical principle of Minimization developed by Leibniz4. This kind of optimization has been used on physical components in the first place but nowadays engineers try to optimize how time is spent.
This optimization affects everyone, and is part of engineers’ role within organizations.
6.3 Engineers and organization
With a growing complexity of systems over the years, organizations have emerged because it was needed to have people work together in an efficient way. This optimization has an impact on every person in the company and even outside. Therefore, engineers have to follow specifications on how to communicate their work. This normalization helps people reduce information loss as well as being more efficient. In fact, due to the entropy of communication5 guidelines and rules are needed to optimize the exchange of information.
Because the structure implemented by the organization gives guidelines and is known by everyone in the company, communication is easier and more efficient. This optimization has some drawbacks as it reduces the time available for innovation which is, as said before, the essence of engineering. This reduction of time for engineers is due to all administrative forms or efforts to fit the normalization imposed by an organization.
Organizations also optimize engineers’ work by limiting resources so that one will think about how to use them efficiently. One just has to make sure that the organization impact on people is not greater than its benefits.
Engineers developed tools, technical solutions, architecture, etc. This conception was static. Now, these tools serve an optimization purpose and a very interesting point is that they can be used for organizations or other dynamic structures. Nevertheless, the result must be discussed. Where one would use yield for systems, efficiency and effectiveness seems more appropriated for people or organizations. The effectiveness represents how good a result with a given set of resources is. The efficiency represents how well resources were used to meet a target. This difference between systems and organizations lies in the precision one can have when comparing the system before and after the modification. A method can be implemented with the purpose of making an organization more efficient but the gain can be hard to evaluate numerically meanwhile the yield of a system is easier to assess.
Engineers’ job has been shaped over the years by their environment. Organizations were needed to make people work together in order to develop complex systems thus, engineers had to spend time to fit organizations requirements. This reduced the time allocated by engineers to develop new solutions to a problem. Energetic optimization is also time
4 Leibniz, le meilleur des mondes possibles
5 Shannon & Mc Luhan
Provisoire 6. ENGINEERING 12 consuming for engineers and has been their main activity since the steam revolution but it is about to change.
6.4 From optimization to innovation
Due to a globalization of the market and a need for industries to always provide better products to their customers, engineers spend most of their time optimizing the energetic yield of systems. They develop solutions and implement modifications to reach the highest energetic yield possible. Nevertheless, engineers, when working in R&D6 departments on pre-design and design phases, spend their time innovating and looking for new solutions to solve a problem. The problem is that over the years the amount of systems needing an optimization has increased a lot. Due to systems complexity, engineers need to abide by organization laws to ensure a good communication. Therefore, the time left for innovation has been reduced for engineers as they focused on optimization.
Innovation is a complex subject; one should not consider that it is reduced to breakthrough innovations. An innovation is rarely a breakthrough innovation such as the wheel, the steam machine or the Internet which are essential to create and open new sectors of activity as well as being a revolution for the society. Most innovations concern small pieces of a system but can lead to a major increase of its energetic yield. Industrials have thus been innovating all the time even if only major breakthrough innovations might have been noticed by people outside of the industry.
Nowadays, with the development of tools such as AI7, big data and knowledge management, engineers are offered new ways to deal with optimization problems. With machine learning and deep learning, optimization algorithms are able to optimize systems faster than a person can do it. People are thus more and more monitoring results given by these algorithms. This is a major change in engineers’ job. Usually this monitoring phase is less time consuming than a usual optimization phase and therefore, it allows engineers to focus on innovation because they have more time for it.
Engineers’ job is about to change drastically and they can now go back to their core activity:
innovation. In fact, they can allocate more time to it due to the reduction of load on optimization problems thanks to the development of AI, big data and machine learning. The major point here is that engineers should use their time for innovation topics rather than being reduced only to monitoring results coming from optimization software.
6 Research and development
7 Artificial Intelligence
Provisoire 6. ENGINEERING 13
6.5 Summary
The essence of engineering is ingenuity. Engineers’ core activity is therefore to develop solutions to a problem as they want to offer a rational answer to a need. Over the years, the need for engineers has been limited but it has increased drastically with industrial revolutions and the globalization of the market. For several decades, this need shaped engineers’ work within companies as they used to spend most of their time doing energetic optimization. The lack of engineers led to a focus on energetic problems and optimization.
Furthermore, with systems being more and more complex, organizations emerged in order to make people work together and be able to address such complex problems. Nowadays AI is reducing the load on engineers concerning optimization problems. Hence, engineers can go back to their core activity and focus on innovation.
One has to keep in mind that economy principles allow people to increase their efficiency but it is hard to evaluate this gain numerically. Hence, yield will only be used for systems while one should talk about the efficiency of a person, of an organization or model development. As it cannot be numbered, efficiency is esteemed by experts that evaluate the impact of a method on a given topic (for instance modeling development).
Even though ingenuity as well as engineers’ job needed to be introduced, this document focuses on economy principles and methodologies used in the industry when developing a model which will later be illustrated by a model evaluating the impact of an organization on aircraft support performances.
Provisoire 7. MODEL DEVELOPMENT 14
7. M ODEL DEVELOPMENT
This section expands the understanding of a modeling methodology used in the industry
7.1 Development cycle of a model
When developing a model for an R&D topic, a generic cycle appears. This development goes through different phases as follow:
Opening phase: the developing team gets to know the subject. It is a short phase.
Exploration phase: in this phase, one takes time to widen his vision in order to make sure the model at least include all primordial elements of the problem and if possible secondary elements to increase the level of details. It is a flourishing phase where one will harvest all ideas.
Closing phase: transition between exploration and expertise phases.
Expertise phase: one will choose and reject solutions while providing an explanation for each choice. This phase usually begins with the association of different solutions or ideas together. A critic vision is needed here. Further details are given afterwards in section 7.4.
Implementation phase: the modeling work will be achieved here to end up with a final version of a model. It can be subdivided between phases where the developing team chooses to really develop the model and thinking phases that can be backed up by sensitivity analysis to choose how to allocate resources for the development.
2nd opening: transition towards another exploration phase.
2nd exploration phase: this phase can be used to think about the refinement of a part of a model or a new part of it that was not in the former requirements. That phase can also be used to think about how to communicate about the model.
Now that one has a general idea of the development cycle of a model in R&D, the next section will deal with its area: whether or not it is interesting to address a specific problem, only dealing with a restricted part of the whole system.
7.2 Models and data structure
When developing models for a complex system, one usually comes up with two different options. One can either choose to use several specific models and put them together thanks to a generic data structure or one can choose to implement a global model. In both cases, one needs to be able to tackle any question concerning the subject represented by the model. Using specific models regrouped through a generic data structure is one way to implement rationalization in model development as it will be presented.
Provisoire 7. MODEL DEVELOPMENT 15 A model regroups the concept, the simulation, different parameters and their value needed to solve a problem. A data structure is a structure that harvest and organize data that will be needed for a model to work. A generic data structure offers enough categories so that every data can be classified in it. Thus, any model integrated within this data structure is able to access any data needed for it to work.
The structure described here helps articulate specific models together and can be seen as a meta-model that represents the organization of models. Such a generic data structure that put together specific models in an efficient way help enhanced models purposes among which:
Helping people to negotiate by giving them results that can back up their arguments
Helping to compute solutions by giving an easy access to the right model
Helping to shine both at an internal (within the company) or an external (with clients, suppliers or partners) level.
7.2.1 Global models
A global model is the first solution that comes up to solve a problem when the area of expertise is changing over time. It usually starts with a simple approach that only tackles a specific problem but a new part is added to the model to answer another question again and again. In the end, one obtains a global model that can answer all questions asked. The data structure and the model are mixed together. Global models can be illustrated as follow:
Figure 1: Global model
The main problem with global models is their size. When a model is further developed to answer a lot of questions, it has all chances to become unstable and get away from a user- friendly model because of all parameters that will be needed for it to work. The model will still be coherent because there is only one entity but the stability is not guaranteed. It might seems easier to use a global model rather than specific ones when developing an activity but one needs to have another solution for a complex and diversified activity.
7.2.2 Specific models
Specific models are developed to answer a specific question. The main difference with global models is that when another question comes up, another model is developed. In order to tackle several problems and questions, one needs to organize model interactions
Provisoire 7. MODEL DEVELOPMENT 16 thanks to a generic data structure. This structure contains all data needed to answer all questions but only uses specifics models that can answer a question without soliciting others.
The main problem with specific models is that they become incoherent if left alone. The data structure that holds them together is designed to ensure the coherency of the whole.
The following schema illustrates that.
Figure 2: Specific models
The data structure should be able to differentiate inputs from outputs and all specific models should be able to interact with appropriate data. Meaning these models will only access needed inputs and communicate outputs that have been evaluated.
Specific models are more accurate and stable than global ones because they deal with smaller problems. They also are more user-friendly by being easy to understand by the end user. Since they tackle very specific problems they only need a reasonable amount of parameters to work. Though, it does not mean all models are simple.
The specificity of a model can be evaluated by the amount of knowledge needed to understand it. Therefore, a model is specific when it addresses a small number of domains, usually one. For instance a model dealing with the maintenance of pieces of equipment can be complex because it covers numerous cycles and pieces but easy to understand because it is dedicated to a specific business.
One drawback with this method would be the number of models that would need to be developed in order for the company to have enough of them to cover all phases of a project and answer all questions. Developing a generic data structure for models also needs a lot of thinking and rigor. Nevertheless, it is also one of the last steps to take for industrials as they developed models throughout their history to bring their products and solutions to market.
Provisoire 7. MODEL DEVELOPMENT 17 7.2.3 Development and update
In both cases, no matter which structure is selected between a global model and several specific models, it was already said that a solution should address all problems. One has to make sure that this requirement is always fulfilled, even after years. That means one has to look at new questions that should be answered or tackle older questions through a new approach. Therefore, an updating of the chosen data structure is mandatory.
Norms are developed to bring some constants. They are unchanged for a few years making it easier for industrials to know which rules they must respect when designing a product.
Nevertheless, over history, norms have changed a lot and will continue to change. When developing a structure, one thus needs to keep enough flexibility so that this structure will be able to fit future requirements and norms.
7.2.4 Data specificities between engineers and operators
When developing models, engineers usually use mathematical laws to characterize functions. However, because they do not deal with such concepts as often as them, operators use observable quantities instead. Therefore, when collecting data in order to integrate it to the data structure, engineers adapt their questions to fit the language of operators. Here is an illustration, instead of using a normal law, they will only talk about the main modes of the function. It means they will only speak about values that are often observed.
Figure 3: Exchange of data between engineers and operators
Now that the choice of structure for models has been studied, different methods to adopt during an exploration phase will be introduced.
Provisoire 7. MODEL DEVELOPMENT 18
7.3 Deductive and inductive methods
When analyzing a situation, there are usually two ways to do so, they depends on the starting point which can be very detailed or much more general:
A deductive method, also called “top-down”, identifies a problem at a macro or business level and then adds more and more details to go to a specific technical solution.
An inductive method, also called “bottom-up”, founds its reasoning on a precise technical solution and then tries to use abstraction to reach a more general situation that suits a business level.
For instance when modeling the impact of an organization on the dispatch of a plane:
Using a deductive method, one starts with general needs and deduces functions to fulfill.
Using an inductive method, one looks at a specific organization in a specific location and extracts different functions that are realized by it.
Functions here represent keys missions or roles that together allow people to do an activity.
They are divisions of actions that need to be accomplished to realize something.
An organization regroups all people working for a company together as well as rules that help to make them work together.
The following figure represents this difference in terms of vision and abstraction. One can either start from top or bottom to end up at the center with a functional vision. This vision helps people to compare requirements at a macro-level and specific answers put in place by an organization locally.
Figure 4: Different levels of vision
Provisoire 7. MODEL DEVELOPMENT 19 The bottom of this structure is a constructional vision that is always changing meanwhile the top of it is an operational vision that is long lasting. One can visualize an interface between the top and the bottom that the functional vision represents with its advantages concerning discussions between people at the business or operational level and those at the constructional level. One must notice that the different in terms of vision is due to the difference between populations at the top or the bottom of the pyramid that do not have the same activities.
When models are developed for an optimization purpose, it means that a solution is already in place and is carried out by an organization. Hence, it is easy to have a description of such an organization and extract functions from it. Nevertheless, one must be aware that organizations are always changing over time and strongly depends on their environment.
Therefore, the extraction of functions from them is not so simple.
An operational vision brings more stability than a constructional one. However, the constructional level regroups more details. As a consequence, one needs to make sure one will be able to break down operational requirements in different functions. In fact, organizations allocate resources to fulfill a function but cannot estimate resources needed and how to organize people if there is too much abstraction and the structure is not divided enough.
Functions are most likely to be unchanged over time while an organization is different for each base and even evolves over time. A parallel can be drawn here with theatre. One would need some roles such as a hero, a main opponent, a confident (to let the public ear internal thoughts) etc. and will affect these roles to each person to end up with a play. Each play is different but functions are the same. The fact that organizations change in time can be seen as the adaption made by a director.
Now that methods for an exploration phase have been presented, tools and methods for an expertise phase will be presented to judge which ideas need to be represented in a model.
7.4 Expertise phase
This phase is essential as it is a step between a flourishing exploration phase, where a lot of ideas are harvested, and an implementation phase, where the model is developed. It is essential that one takes time to sort out these ideas, make a clear selection and a list of what need to be realized during the implementation phase that will follow. This is a crucial step towards a rationalized model development.
7.4.1 Doctrine, also named philosophy or principles
A doctrine is the first thing that should be set when working. It helps people to make decisions and judgements especially during an expertise phase. It is needed throughout the entire model development but crucial during this phase.
A doctrine characterizes an environment and also decisions that needs to be made concerning different situations: for instance, with whom to work together? Choices made by a doctrine are at the business or operational level and set major directions and rules on how to conduct a project.
Provisoire 7. MODEL DEVELOPMENT 20 Situations can be:
Rewarding: one will tend to maximize this situation.
Acceptable: this situation is doing no harm but does not bring any additional value.
Annoying: this situation has to be reduced if possible as it brings a small negative value to a system.
Unacceptable: this situation must not happen.
A technical framework evaluates situations in the same way and results from an adopted doctrine. A simple illustration for that can be as follow:
System Real state
Functional Not functional
Estimated state Functional Rewarding Unacceptable
Not functional Annoying Acceptable
This illustration is voluntarily very general and only serves to illustrate the theory. For the model developed during the internship, decisions were made by classifying data as explained in section 7.4.3
Being able to classify ideas and potential solutions with this system can be an efficient way to judge if they have an interest being implemented in the model.
7.4.2 Engineering processes
A process represents all activities that contribute to a specific goal of an organization. It can be part of a working phase or cover several of them. Military support processes are listed and detailed in section 8.2.1.
When developing a model and in order to be as streamline as possible, one should be well aware of the granularity of a model. It represents the level of details in a model and is not the same according to the process a model interacts with. Therefore, processes have a key role during an expertise phase as they set boundaries. In fact, processes mainly set the level of details needed or available in a model. Someone working in a pre-design phase where a system is only defined with main concepts will not be as detailed as one modeling an in-service phase where a system is operational and all details and every piece of equipment is known.
One can see that the granularity of a model is easy to evaluate when it only interacts with one process. Though, it may be more difficult for models that are transversal and cover several processes at the same time. This point is again about global and specific models. It is better to have several models that deal with one process at a time than a global one that interacts with many of them so that a single level of granularity can be chosen for the model and one does not have to think whether or not to refine the model specifically on one topic.
Provisoire 7. MODEL DEVELOPMENT 21 Concerning processes, a decision can be made whether to create a model during the pre- design phase and to refine it along the way to meet the precision level required by other processes or to create several models, one for each process.
7.4.3 Classification of data
When developing a model, one need to identify elements or characteristics that are essential and those that are not because resources are limited and everything cannot be represented in a model. Each elements and characteristics has to be looked at and only elements with a big impact on the system will be modeled meanwhile the influence of others can be neglected.
Therefore, the following hierarchy can be used to classify elements and characteristics:
Essential elements
1st order of importance: Even though all elements are needed for a system to function, a decision can be made when modelling it to only represent essential elements of the first order of importance.
2nd order of importance: The difference between elements of first and second order of importance is that the contribution of second elements can be neglected. This contribution shall fall into an acceptable range of error defined in advance. If the contribution of a secondary element exceeds this admissible error it means it should be consider as an element of first order of importance.
Incidental elements
For a system some characteristics are essential while others are incidental. These last characteristics have a limited impact on the final system. They are a result of a decision made at the time, sometimes due to limited technologies. For instance, when humans wanted to fly, they studied birds. Engineers finally discovered that the shape of a wing is essential meanwhile the material (feathers for birds) is incidental which allowed humans to use wood and fabric in the beginning while they use aluminum or composite based materials now.
One must realize that the decision made to classify an element or characteristic essential, of first or second order of importance, or incidental is very subjective. A developer can consider a point incidental meanwhile it is essential from the client’s point of view. For instance, the color of a pull is incidental for the industrial making it when considering the taint cost is almost the same for main colors. On the contrary, the sales director will consider color essential as black pulls sell way better than yellow ones for instance. An essential element of second order of importance can change and suddenly become of first order of importance when its impact on the final result grows.
For instance in the model developed during the internship, the contribution of pieces of equipment with a good reliability to a potential failure was first estimated of second order of importance. After a sensitivity analysis on this parameter, people realized that having too many of them was as bad as having one element with a bad reliability. It means that having ten elements with an “average” reliability would cause as much failure as one element with a “bad” reliability. Hence, all elements were considered according to their reliability and the number of elements with that given reliability.
Provisoire 7. MODEL DEVELOPMENT 22 The key point here is that not everything can be modeled due to cost considerations and that the level of importance of a characteristic strongly depends on the point of view.
Therefore, a good communication should be established between the developing team and the final customer.
7.4.4 When harmonization helps decision making
During WWII, General Eisenhower was appointed leader of the Allies forces and had to make decisive decisions on how to manage the war. The situation was very complex and he was not able to make a decision easily because he was presented a lot of different scenarios. Each scenario came with different parameters and explanations, therefore the decision making was even more complex.
A decision was taken to harmonize the way scenarios were presented to headquarters.
People were thus asked to present a scenario with a limited number of parameters that were always the same. For instance, the cost of scenario was presented according to the amount of resources (coal, wood, food…) consumed. Thanks to this new approach, Eisenhower’s administration could choose between different scenarios far more easily.
One can see that this rationalization illustrates the idea that too many elements and too many versions are slowing or even blocking the decision process at the operational level.
Everything could not be understood so only elements that affected the result deeply were considered.
Nowadays, systems become more and more complex so that one cannot understand a whole system with all details. Either one can be an expert and understand a part of a system at a very detailed level or one can understand a whole system without too many details being a generalist.
One can see here that rationalization processes started years ago to harmonize and simplify model development. This work is still needed today due to the complexity of systems. Being able to take decisions based on an harmonized description of a project greatly helped engineers and decisions makers.
7.5 Sensitivity analyses
When in the implementation phase, one can go through multiple cycles in order to progress step by step. In order to choose which actions to carry on during the next cycle sensitivity analyses can be used. A sensitivity analysis evaluates the impact of a variation of a parameter value on the final result. This is very useful when refining a model and one need to know which parameter has the most contribution to the final result, in order to make sure it is detailed enough. Variations of parameters values can be problematic as a parameter that has a small contribution to the final result might be a main contributor with another value.
The idea with sensitivity analyses is to govern the development of a model so that it will be robust. In fact, the refinement in the model conducted after a sensitivity analysis ensures that the end-user will have a smooth observation of results.
Provisoire 7. MODEL DEVELOPMENT 23 Sensitivity analyses also help to rationalize the development of a model by orienting resources where they can be used in an efficient way. The refinement implemented will bring more value for a selected parameter than a random one. Hence, the value of the model will be greater with less resource spent. It also avoids waste of resources by not giving possibilities to refine a part of a model that is already detailed enough.
7.6 Incremental development
7.6.1 Concept
Incremental development is a modeling methodology used when one wants to model complex systems. It is based on a strong assumption: the possibility of modifying a part of a model without impacting the rest of it. Therefore, incremental development allows a development of a model deeper and deeper. Each step is often followed by a sensitivity analysis in order to orient the work for the next step in the direction that will produce the biggest additional value.
Incremental development fits most situations and offer huge advantages. This method is really useful as it breaks down a complex system into simpler pieces. Thereby, one can understand a complex system as a whole because all detailed information has been regrouped into some appropriate categories that easily represent the whole system. One can still access detailed information of a specific part of a system for an expert analysis.
One can also only get the essence of each part of a system to get a global vision of it.
When developing a model using this method, one will first model the macro level that represents the main capacities and parts expected for a system. Each description will be voluntarily gross in the first place so that one can discuss and explain each part of a system in a general way. Multiple cycles will then follow to refine the model in order to fit customer’s expectations.
7.6.2 Pros & cons
Incremental development has many advantages, the main one being the capacity to break complex systems into understandable pieces. Incremental development also contributes to growth potential. That means that the primary gross approach and the development directed by a sensitivity analysis leaves part of the model in their raw state. This is a very important point, one can argue that details are missing in the model but one should think about possibilities of such a decision. By modeling small pieces of a model at a time, one makes sure that a developer will go from scratch to a finished part all on its own. This increases drastically the global efficiency as continuing someone else’s model is way harder than modeling something all by oneself.
Using a method such as incremental development is quite effective for humans as they now have modeling tools that allow them to model a very wide range of characteristics. In fact, one can now model multiple physical aspects of a system. For a plane it goes from the shape, the aerodynamic, the structure to the thermodynamics of engines. One needs to know which directions will be needed according to the activity. Furthermore, humans are
Provisoire 7. MODEL DEVELOPMENT 24 excessive by nature and this might push them towards an ineffective modeling. By limiting resources and only giving a limited part of the model to a developer, one prevents ineffective work thus increasing the efficiency of model development. Thereby the structure of incremental development gives some rigor and some restraint and allows a better allocation of resources for project development.
Even though incremental development has many advantages, it also has some limits. For incremental development, combinatory problems can be an issue: a small change can have a big impact on the final result. In fact, the main assumption of incremental development is that the refinement of one part of the model does not impact other parts. This ensures models are robust as long the new refined part is as robust as the old one. An issue is that with combinatory problems, one need to reevaluate the whole model when refining a part of it because a slight modification might impacts other parts.
7.6.3 Summary
Incremental development allows people to model very complex systems by breaking them into understandable parts. It has been developed during the 20th century and has some limits with combinatory problems. This method has been used to carry out modeling tasks during the internship because it has been ensured that it can be applied for such problems.
This method allow people to develop models in a streamline way but one should keep in mind that a model alone is not efficient for an organization and it has to be implemented into a structure that governs and organize them.
7.7 Industry specificities
7.7.1 Constraints, among which the limitation of resources
Because industrials evolve in a real world, they only have access to limited resources. This is one of the biggest differences between theory and reality. Resources represent iron, copper, money, time or humans through their availability, competences, etc.
Industrials have an interest at minimizing the waste of any valuable resource, therefore they have to think about how to use them in the most efficient way possible. This point is the main motivation for all techniques and methods presented before and used by industrials.
Industrials have to find a compromise between cost, quality and delay for a system or a project. The final goal cannot be reached if there are too much losses or an inefficient use of resources during the development.
When developing models, one reaches better results when alternating between reflection and implementation phases rather than mixing them. Dealing with a small topic is also better than dealing with a big one as one can focus on a specific part rather than trying to develop all of them together. Of course, industrials need to tackle wide subjects and will therefore develop several parts of a system at the same time but they will affect each part to a specific team that will focus only on a single part.
Provisoire 7. MODEL DEVELOPMENT 25 As companies are self-funded, their best option to increase the amount of limited resources they have is to make profit on their systems or solutions. One has to think about the profitability of a system. That means an industrial should not spend resources if the benefit is not greater than the cost. In order to progress and continue to deliver products to market, industrials need to lower their development cost by rationalizing their use of resources.
7.7.2 Competitive global market
Most of the time, companies evolve in a competitive environment which makes them more dynamic but also more vulnerable as their competitors are willing to rule the market and increase their share or it. In fact, due to the globalization of the market, one must consider that competitors are always innovating and it helps to maintain efforts towards innovative systems.
Because offers available on the market are always increasing, industrials need to provide systems that integrate the sum of all innovations. One major point here is to notice that very few customers are ready to spend a higher price for that, they require such a product for the same price as the previous product.
Within such economic environment, the survival of an industry is at stake. The price to pay is a mandatory improvement in resource management.
7.8 Summary
Optimizing, improving and innovating are more than technical satisfactions. The survival of any industry is at stake, meaning a financial and social responsibility.
If the quality of the modeling process is not the only factor of success it is a crucial tool for optimizing engineering work.
In the next chapter the way of which the above principles have been implemented will be presented along with the concrete example developed during the internship. It will both serve as an illustration of theories presented here as well as presenting the impact of an organization on support performances.
Provisoire 8. INTERNSHIP 26
8. I NTERNSHIP
8.1 Introduction
This section regroups an introduction to the specific context of Dassault Aviation, the department I have been working with, its activity and the purpose of the internship. The main focus here is on information that brings added value for the understanding of the subject of the internship. Further details are available in Annexe C.
8.1.1 Dassault Aviation
Dassault Aviation is a major actor in the aerospace industry, both civilian and military. This dominance benefits from an ability to manage both business and military requirements for decades. Moreover, Dassault Aviation is the only company that designs, manufactures and supports both business jets (tools for work and economic development) and military aircrafts (instruments of political independence). It is also the last company owned by the family of its founder: Marcel Dassault. The company can also count on more than a century of history to accomplish its key role for the national sovereignty.
To stay a leader in this industry, Dassault Aviation is always evolving and looking for new tools that can enhance its products and widen its offer. Therefore, the company developed the Rafale, an omnirole8 fighter that replaced seven different aircrafts and is proof of a major optimization result. This aircraft is combat proven, has been in service for many years and is still one of the best military aircrafts as its different versions keep it equipped with top-end technologies.
Figure 5: Rafale
8 An omnirole aircraft can accomplish every role it is design for during the same flight and simultaneously in opposition to a multirole aircraft that can accomplish different roles but with different flight configurations.
Provisoire 8. INTERNSHIP 27 Over the past few years and thanks to its experience and knowledge, Dassault Aviation has been leading the nEURON project in Europe. It has been the main architect, coaching and leading several companies in order to work together. This UCAV9 has been developed as a demonstrator and is a successful partnership between European leaders of the aerospace industry.
Figure 6: nEURON
Recently, Dassault Aviation, along with Airbus, has presented the FCAS10 project for the future of military aircrafts. This project has been approved by France, Germany and Spain and will be carried out by several Europeans aeronautics leader together to achieve the project.
Figure 7: New Generation Fighter
9 Unmanned Combat Air Vehicle
10 Future Combat Air System
Provisoire 8. INTERNSHIP 28 In the meantime, Dassault Aviation is always developing its offer concerning business jets.
The latest project for that is the Falcon 6X which will provide the largest cabin for business jets. Moreover, Falcon aircrafts are designed to provide top-end comfort to their passengers and offer the largest operating distance possible while being more fuel-efficient therefore being more eco-friendly. Here is a picture of the latest Falcon released, the Falcon 8X.
Figure 8: Falcon 8X
The dominance of Dassault Aviation also comes from reforms in factories. These reforms aim to provide the newest manufacturing capabilities in order to increase yield while the organization always evolve to increase the efficiency. These reforms also come with a reduction of the impact on the environment due to production. In ten years, the amount of water used has been reduced by 50% and the amount of volatile chemicals by 57%. The amount of waste has been reduced by 40% while the valorization of remaining waste always increases over the same period, reaching over 90% for non-hazardous products and over 50% for hazardous ones.
Finally, engineers within Dassault Aviation are offered new tools to enhance their models even more and thus their services for customers.
8.1.2 DGSM11 / DIS12
Dassault Aviation as an aircraft manufacturer provides support services for customers. This fact is important as the technology equipped in both military and civil aircraft is becoming more and more complex and costly. Moreover, this cost increases with the percentage of software and firmware in the system.
Roughly speaking, the cost of support for an aircraft over its life is equal to the price of the plane itself, which is why optimization is essential. Furthermore, this cost can increase even more according to the activity planned. Having multiple planes ready on several bases for different missions is more costly than preparing one plane for the said mission. However,
11 Direction Générale du Soutien Militaire = Military support general direction
12 Direction de l’Ingénierie du Soutien = Engineering support direction
Provisoire 8. INTERNSHIP 29 the preparation time is sometimes so small that militaries need to have planes ready before the mission is announced, for example for air regulation or interception missions.
Dassault Aviation is divided into several branches called “directions” and subdivided in departments and services. DGSM represents the support branch and DIS is the department within it in charge of engineering support.
This organization has been made so that a branch of the company (DGSM) is dedicated to aircraft support. This is because customer’s issues need to be tackled as fast and efficiently as possible.
Among other activities, engineers within the DIS department develop cost and availability models for current and future aircrafts. At the moment, the whole department is in an on- going reform to develop a structure that can easily articulate existing models. This reforms aims to offer an easier access to a specific model that fits engineers’ requirements for a specific study. This reform motivates the characterization structure developed by the author to address a model.
Most of DGSM activity is directed towards military customers as they are more willing to buy and maintain a whole fleet (mainly for being autonomous in times of war) whereas business jet customers often have a single or limited number of aircrafts and tend to do the maintenance at big airliners support stations. Support logics and considerations are radically different between civil and military sectors. Nevertheless, even though civil customers often do not perform maintenance on their aircraft themselves, Dassault Aviation developed the Falcon Response to provide a fast assistance in case of a failure so that the impact for customers will be minimal.
8.1.3 Aircraft support
Even though aircrafts and especially military aircrafts are complex systems, one needs to be able to perform maintenance in an easy way. Thus a huge work has been accomplished by aircrafts manufacturers to enable a fast and easy repair. Every piece of equipment is divided into small parts which are most of the time plug and play. This means that one can access the equipment and replace it or part of it quickly by removing the damaged component and replacing it with another functioning component. The piece of equipment is replaced and the damaged part then enters a repair process that will not be presented in this document. Manufacturers also developed an exhaustive documentation in order to offer guidance for customers when repairing an aircraft or for the maintenance plan that one must follow to keep his aircraft navigable.
Aircraft maintenance also has to abide by many different regulations:
Navigability: one has to prove that the plane it is flying with is an exact copy of an authorized model by regulators. One thus needs to maintain the plane as the manufacturer recommended and then present proves to be able to fly.
REACH13 regulation from the EU14: since 2007, EU has put some rules and laws in place to protect the population from an intensive and unproportioned use of chemicals by
13 Registration, Evaluation, Authorization and Restriction of Chemicals
Provisoire 8. INTERNSHIP 30 industrials. For instance, use of chromates to prevent corrosion on planes is now forbidden.
Polluting emissions:
Greenhouses gases: with an increasing market, the aeronautics is under a lot of pressure to reduce its footprint and therefore engagements were taken years ago by all main manufacturers to limit their emissions even though the market is developing.
When maintaining an aircraft, one has to make sure that every system is working to ensure the plane respects its level of emission.
Noise: Especially during take-off, planes are very noisy. Because most airports are surrounded by habitations the level of noise has to be monitored. A good maintenance ensures that noise emissions respect these norms.
8.1.4 Motivation of the internship
Currently the company is working and solving problems addressed by customers in a very good way and with good satisfaction. As the aeronautics sector is very competitive, Dassault Aviation is taking opportunities on new modeling technologies in order to stay a leader and continue to provide top-tier support assistance to their customer. The company is investigating new technologies such as big data, virtual reality or additive manufacturing in order to keep delivering the best products possible to customers.
As a consequence, a reform started years ago and is still ongoing. Its goal is to create a structure, mainly for data, that will help to articulate models together in order to provide a better access for engineers. This easier access enables a more efficient work, thus enhancing response to customers’ needs.
The work developed during the internship and presented here after contributes to this reform conducted by the company. First the data structure developed to characterize models will be presented, followed by a model that aims to model the impact of organization of support performances.
8.2 Characterization of a model
8.2.1 Structure
During the internship a structure has been developed in order to address different key questions that should be discussed about a model. These questions should be answered when developing a model and resulting answers should help a user understand how a model works and in which context.
This structure tackles quantifications models but is voluntarily general enough for other groups of models. Several groups were listed such as:
Requirements (submitted by customers to Dassault Aviation)
14 European Union
Provisoire 8. INTERNSHIP 31
Specifications (transmitted by Dassault Aviation to suppliers)
Descriptive models (that serve for illustration purposes)
Quantification models, calculation models (giving results)
Simulations (that are dynamic models)
Database
This characterization structure helps to ensure coherency between models by giving a quick and easy description of a model. It also helps to detect redundant models in order to select the most appropriate one and remove others. This structure has been divided in different points as follow:
Uses: A model can have different uses according to the user. A result produced by the model can be used in different ways by the end-user. For instance calculations results can be used to choose between different scenarios or can back-up an explanation about a scenario therefore defending one scenario without giving information on others.
Phases and processes: The industry divides a project into several phases and processes. Processes can belong to one or several phases. This section addresses the process a model should be used into: for instance pre-concept or in-service phase. It determines when a model shall be used. A model might be used in different processes.
Though, some processes are very different and require different models. Processes for aircraft support are represented in the following schema:
Figure 9: Phases and processes for aircraft support
