Supporting Demand Supply Network Optimization with Petri Nets - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 380

Table 1. Sample response times of Petri Net tool

Number of DSN Analysis time (sec)

Alternatives

0.

0

0 0

to determination of variant effect on the optimal demand supply network setup can be found in (Tynjälä
& Eloranta, 2007).
The typical performance of the tool is shown in Table 1. These measurements were single data points
taken during a typical working day. The response times of the tool are practically the same globally, as
the transport delays are small compared to the analysis time.

discussion and future work

Applying Petri Nets and reachability analysis to demand supply network optimization has the benefit
of exploring the complete state space of options. In wake of risks on the demand supply network, it is
vital to explore also the expensive options for future contingencies.
I developed a High Level Petri Net model expanding on the AND-AND, AND-XOR formalisms
presented in (Baer, 1973; Agerwala, 1973), along with the associated reachability analysis algorithm.
The Petri Net model is general and it may be employed in various types of businesses. In terms of ex-
pressive power, DSNnet stays within the normal Petri Net confines.
A tool based on the DSNnet methodology was developed during 2004 inside Nokia Corporation,
and it has been in active use since February 2005. The tool has been used in all business units and the
results are very positive. The users have especially thanked the possibility of designing the product
structure together with its demand supply network, which was not possible with the former spread-
sheet methods. The work relating to DSNnet tool inside Nokia is ongoing, and the second release of the
tool became operational in October 2006. This release enhanced the tool with multimodal transport
capabilities and improved duty and tax considerations. In the future, the tool is envisioned to become
integrated to a data warehouse, which updates the cost and transport parameters automatically based
on actual business transactions.
The current method may be used in aiding hierarchical “zoom-and-focus” analyses on demand supply
networks. Simulation-optimization suffers from the fact that the modeling languages used in optimiza-
tion and simulation are very different. This means that an off-the-shelf optimization engine is unable
to directly export models to an arbitrary discrete simulation engine. Petri nets are a way of making
transition from static to dynamic analysis realms straightforward (van der Aalst et al., 2000; Billington
et al., 2003). An interface from ExSpect analysis tool to other Petri net simulation tools was easy to
construct (van der Aalst et al., 2000; p.461). Moreover a Petri Net Markup Language (PNML) initiative
has been formed to render the transfer of models between any Petri net tools possible (Billington et al.,