For a full description, see: Lindemann, U.; Maurer, M, Braun, T.: Structural Complexity Management. Berlin: Springer 2009, pages 155-170. (with friendly permission by the authors)
The vehicle safety department of an automotive OEM faced the challenge of carrying out a structural analysis of different sub-systems of frontal and lateral safety to acquire the basis for future efficient complexity management. The following list contains the main objectives of the use case described:
- Creation of system transparency and understanding
- Awareness of typical change impacts
- Comprehension of domain-spanning dependencie
- Design of robust product structures concerning product adaptations
- Identification of opportunities and restrictions for product adaptation
The procedure of structural complexity management was implemented for two reasons: to analyze the frontal protective system which focused on OOP (“out of position”: this means that the front-seat passenger is not in the optimal position during a collision); and to analyze a lateral protective system, which focused on understanding the overall system.
In the system definition phase, the domains involved, the elements of the domains, and the types of dependencies within and between the domains, were defined. For this task the Multiple-Domain Matrix (MDM) was applied as storage for information about domains and dependency meanings. The above figure shows the matrix that resulted from the system definition for the case of the lateral protective system. It can clearly be seen that not all subsets of the MDM are occupied. Many dependency meanings between domains do not exist or were not relevant for this examination.
The methodical application of the MDM allowed a significant contribution to be made in order to create a system model as complete as possible. If all matrix subsets sequentially bring all domains in relation to each other systematically, no significant system connections are overlooked, which means that the MDM can serve as a check list for system definition.
The system dependencies that are located in various DSM and DMM subsets in the MDM were collected in several workshops through interviews with the appropriate technical experts. The acquisition of the dependencies was first carried out for the DSMs along the diagonal of the MDM and subsequently for the DMMs.
The data collection and analysis processes were carried out using the software LOOMEO. LOOMEO allows checking for possible indirect dependencies during the acquisition phase. Plausibility analyses were carried out after each workshop. Discrepancies and conspicuous features in the structures were analyzed and brought forward at the beginning of the following acquisition workshop to be discussed in detail and incorporated into the system model.
Discussions often came up during the acquisition of dependencies, and the results were documented as accurately as possible in the form of comments. It is by this means that an extremely valuable knowledge base was created.
Deduction of indirect dependencies
The network of indirectly linked people was deduced based on the dependencies within components and between components and people. This means that two people are linked, because both address (different) physical components, which depend on each other.
In the example shown in next figure, both person A and person B are responsible for separate components, as can be understood from the information of the DMM; person A is responsible for component 1, which has a change dependency on component 2 (depicted in the DSM); this component is in the realm of responsibility of person B. Consequently, a dependency between person A and person B can be derived, because adaptations executed by person A affect person B by the indirect linking of the components.
Product design application: Improved system management
Different methods were applied to derive handling instructions for improved system management in product design:
Impact check list
The transparency of consequences due to changes in the system was created for various elements by impact analyses. An impact analysis serves to represent the impact resulting from direct dependencies and dependency chains due to adaptations to the system. The impact check list represents a methodical analysis providing information about the nodes directly linked to one specific node in question. The provision of this information supports the systematic step-by-step evaluation of impact propagation resulting from the adaptation of a specific system element. If an adaptation has to be executed, product developers can sequentially evaluate the probability of change impacts to further elements. The systematic analysis of dependent elements guarantees that no possible impacts are neglected.
In the present use case, the impact check list comprises all nodes connected by outgoing dependencies only. In each case, the starting point is the element which has to be adapted. The structural environment is then modeled using this element as a starting point. Outgoing linkages of the elements are tracked so that elements that are directly affected can be identified. After these affected elements have been evaluated, the next level of dependencies is focused on until the entire network is processed.
One of several striking questions in the underlying use case of the front protective system was the question about the impact resulting to the system because of the weight reduction of the inflator. Such a weight reduction could be realized by a reduction of the wall thickness of the inflator – a bottle storing a mixture of argon and helium gas. It was known that by keeping the inflation pressure unchanged and increasing the inner volume of the bottle, the reduced wall thickness would lead to a change of the gas mass flow curve progression of the inflator.
To analyze the impact of this change on the entire system, the active locality of the element mass flow curve progression was modeled in the network. The following figure 3 shows an impact check list for the system element “Mass flow curve progression”
To evaluate the severity of impact, the elements influenced by the mass flow curve progression were colored according to their activity value. The figures above and below show highly active elements (colored red) that could possibly be affected by the planned change. Due to their high activity values these elements can spread changes to a multitude of further elements in the considered system. Therefore, in the next step these elements were investigated in detail. The effects of such system changes were generally unknown before executing the analysis and measures of change management, which relied mainly on the knowledge and experience of several experts. With the impact analysis on hand, an overview of the effects could be conducted in a systematic and time-saving manner.
The checklist-like character of the impact analysis ensures that design decisions are performed and completed at an extremely high level of proficiency. This supports change management in a very efficient way. The developer obtains a deeper understanding of the possible consequences when implementing a change. In particular, it opens up the possibility of systematically identifying the impacts of changes of critical elements. A conclusion can be quickly reached as to whether a change is highly critical or not (and required resources, for example, can then be better planned).
The trace-back analysis was adopted to identify the causes of existing problems. Starting from the last element influenced in a dependency chain, a reverse analysis goes backwards step by step through the chain of impact. Elements that remain constant can be hidden from view. A summary of possible problem causes is finally obtained.
A decisive question in the use case of the lateral protection system concerned the causes of problems with the criterion rib deflection. Therefore, the trace-back analysis was carried out for the parameter rib deflection. As depicted in the next figure, rib deflection is directly influenced by a variety of other elements. This becomes obvious by modeling the passive locality of the element. Furthermore, the influencing parameters themselves are influenced by other elements. The coloring of the elements indicates their passive sum. In this way highly influenced parameters become visible.
With this knowledge on hand it was possible to systematically process the indicated parameters influencing the rib deflection, and the cause of the problem could be rapidly identified in a practical application scenario.
Optimization of the communication network
Another problem represented the question of how dependencies within the component network affect the people network. In other words, how people were linked to each other through their interaction with physical system components was not fully understood. For this reason, a matrix of indirect people interrelations was computed within the framework of the “deduction of indirect dependencies”. The resulting network provided information about the linkages between people due to their work on different but mutually linked components. It indicated that if one developer implements changes to his dedicated component, the second developer has to react, because his component probably needs to be adapted, too.
The figure below shows the computed network of developers as a force-directed graph. It can be seen from the depiction that an area of communication has formed in the center. This can be interpreted to mean that usually communication has to pass via the same subset of developers. This computed network allows one to see how the people structure overlaps the existing organizational structure due to component dependencies. The rearrangement of people in the existing organizational units would be the next issue to investigate. For example, the close and necessary integration of suppliers in the communication process is conspicuous in this case.
The work undertaken led to a considerably increased understanding of the safety system considered. Previously unknown dependencies and correlations became visible and controllable when depicted in this way, and measures for adaptation could be more efficiently managed. This led to an increased reliability in the development processes while simultaneously saving resources.
Improved system design
Besides the methods of controlling complexity, a further step consists of designing the product structures to become more robust in order to reduce the impact of required modifications in general. Several structure attributes (clusters, circular paths, hierarchies, etc.) can be considered in order to achieve this.
Various subsystems are involved in the lateral safety system. The fewer dependencies that exist between the individual subsystems, the fewer impact changes will spread through the whole system. The objective was to generate independent subsystems (to a degree that is possible) in order to design a system that would be more robust in general. Changes would then be allowed to have some impact on the delimited modules; if at all possible, however, the effects should not spread to the entire structure.
The last figure shows the change dependencies between components that play a role in the lateral safety system. The matrix was reordered using cluster algorithms. It shows that various clusters can be identified: seat surface, seat backrest, door and greenhouse, including belt. The clusters are to some degree in mutual dependency.
A significant dependency was identified between the clusters’ seat surface and the greenhouse, including the belt. Here the potential was highlighted to separate the two clusters completely by eliminating only one bidirectional dependency between two elements. The technical solution to eliminate the dependency consisted of integrating the belt connection into the seat.
Use Case provided by Maik Maurer