Challenges in product family knowledge modeling and analysis: from product design to manufacturing


In the automotive industry in general, but in particular for trucks, the vehicle can be configured to fit specific needs of every customer. For trucks, this means that one truck physically can be very different from another truck, for example one truck may be configured to support long-distance transportation, another will be used for inner-city distributions, and a third one will be used for terrain driving. This flexibility puts demands not only on the product development where the mechanical, electrical, and software systems must all interact correctly for all possible configuration it also imposes challenges when it comes to the development of the manufacturing system where the assembly line has to support every possible truck that can be configured. Product families are one way of categorizing products with similar characteristics. A common approach to reduce development costs is to define product platforms with reusable components that can be used for building vehicles for the different product families. Classical feature diagrams are often used to have a compact representation of all possible vehicles for all possible product families. In the talk we will describe some of the challenges with building tools that will support not only the product developers, but also the engineers responsible for developing the manufacturing systems by automatically analyzing the consequences of introducing or removing new features in the product. The problem has a few key characteristics. (i) The number of possible configurations is huge, in the order of 10<sup>100</sup>. (ii) The products have both mechanical parts, electrical parts, and software; each of which is configurable. The configuration of the software will depend on how the mechanical and electrical parts are configured. Variability of mechanical, electrical, and software parts are typically modeled in different tools, but they are highly interdependent. (iii) When assembling a product, many additional constraints must be taken into account. Typically, the assembly of a truck is done in an assembly line divided into different cells, where the truck moves sequentially from one cell to the next. This implies that the work in the different cells has to be distributed evenly between the cells. In addition, we have precedence constraints, expressing in which order different parts may be assembled. This might be due to geometric-feasibility where there has to be a collision-free path to bring two subassemblies together, or stability where two joined parts have to maintain their relative position and not break spontanously, for example, due to gravity. Another complication is that some tasks maybe only done in some cells due to some specific hardware which is needed to complete the task. (iv) A high number of stakeholders are involved with different responsibilities from mechanical engineers, electrical engineers, control engineers, and production engineers but also business people who decide which configurations should be available on the different markets. Every stakeholder has its preferred way of representing and interacting with the relevant information. For example a market analyst will need a textual informal document while a product designer will need a formal graphical document. Thus, a common problem is that often the same information is represented in different IT-support systems. This approach often results in redundant representations. Another problem can be that there are no exact semantics associated with information. A challenge is therefore how to avoid having redundant representations of the same information, but still being able to support different stakeholders? It is also desirable to be able to associate a well-defined semantics to all the knowledge and information in system. Avoiding redundant representations and having clear semantics is a prerequisite for introducing computational methods that will enable automated analysis of consequences of design changes to the product or the manufacturing system. Since exhaustive enumeration of all possible vehicles is not possible, another challenge is the development of algorithms that can evaluate the manufacturability of each possible vehicle. In our work on meeting some of the challenges described above, we are working on integrating domain-specific models of the product family. This semantic model is enhanced using a previously defined meta-modeling language, Clafer, to enable model analysis. The semantic model also describes the assembly tasks, precedence relations between tasks, and capabilities of resources. In order to support different stakeholders and allowing them to work using a representation that is familiar to them, we are building the prototype tool upon the Domain Workbench from Intentional Software. The work is part of the EU FP7 project Know4Car.

DOI: 10.1145/2660190.2662954

Extracted Key Phrases

Cite this paper

@inproceedings{Ebrahimi2014ChallengesIP, title={Challenges in product family knowledge modeling and analysis: from product design to manufacturing}, author={Amir Hossein Ebrahimi and Pierre E. C. Johansson and Knut {\AA}kesson}, booktitle={FOSD@GPCE}, year={2014} }