Collaborative e-work parallelism in supply decisions networks: the chemical dimension

Abstract

In today’s increasingly networked and interconnected environments, business processes and associated decisions tend to span across organizational boundaries, making traditional centralized coordination impractical and/or unfeasible. Furthermore, collaborative networked organizations need to be able to respond to rapidly changing demands and requirements, under uncertain conditions and without centralized control. This situation emphasizes the need for adaptive, distributed, and self-coordinated supply network decisions models. This study focuses on the efficient coordination of parallel, reconfigurable, inter-organizational supply operations. The chemical dimension of collaborative e-work parallelism is introduced, including a novel market-based mechanism that supports supply networks’ parallel and decentralized reconfiguration. The newly developed approach is illustrated by examples from a global industry network, demonstrating its advantages, and identifying its limitations.

Appendix: PRIO simulation model

In order to perform the experiments described in “The parallel reconfigurable inter-organizational (PRIO) model” section, the PRIO model logic was simulated using the Arena simulation software. The PRIO simulation model logic is schematically depicted in Fig. 13.

Fig. 13

PRIO simulation model diagram

Fig. 14

Bid calculation diagram

The historical daily orders placed by customers at the different destinations zones are stored in a database including all order and customer related information. The simulation model reads all the relevant information from de historical file and generates entities representing these orders. Order size, customer ID, geographic zone and corresponding date are stored in the model as entity attributes.

Depending on its size, an order can be labeled as Type 1 or 2. Type 1 orders are those of size near a full truckload-between 55 and 60 units. Type 2 orders are those of smaller size. If orders are bigger than 60 units, the model splits them in as many full truckloads as possible, and the remaining units are assigned a Type 2 attribute.

When orders are the size of a full truckload (Type 1), they can be supplied directly from the manufacturing plant, or through the distribution resources of layers 1 and 2-distribution centers and cross docking stations. However, Type 2 orders cannot be supplied directly from plant, since the manufacturer does not count with the infrastructure required to handle small quantities of products. Accordingly, Type 2 orders can be supplied only from distribution centers and cross docking stations.

The simulation model sorts the different type orders and splits Type 1 orders into three different entities that share the same attributes as the original order. These entities are routed to the logic representing the three supplying alternatives for the order, namely manufacturing plant, distribution centers, and cross docking stations. On the other hand, Type 2 orders split into only 2 entities which are routed to the two supplying alternatives for this type of order-distribution centers and cross docking stations.

The simulation model calculates a bid for every possible route through the distribution network, according to the bidding function introduced in “The chemical dimension of CEP” section. During runtime, performance of individual distribution resources is recorded in order to use that information in the process of bid calculation: a penalty factor is calculated as a function of the difference between the promised and the real generated service time, every time an order is processed by a resource. When the real time is bigger than the promised time, a penalty factor is recorded by the model. That factor is used to adjust the bid of that resource for the next order.

Every block of the Bid calculation section of the diagram is actually a sub-model which presents different levels of complexity depending on the supplying source that the logic describes. Figure 14 shows a diagram describing those sub-models.

Bids are calculated for every alternative route in every resource logic by using the bidding formula and the penalty factor. Costs and promised service times are obtained from an external file containing information regarding every resource. Real times are generated according to triangular probability distributions that have been previously.

Figure 14a shows that every orders supplied directly by manufacturer has one and only one possible route throughout the network, so only one bid is calculated for that case. However, when an order is supplied by a distribution center, there are two alternatives paths through the network, since there are two distribution centers from where that order could be supplied. Accordingly, two bids are calculated, as shown in Fig. 14b.

Finally, the more complex case of orders supplied by cross docking stations is shown in Fig. 14c. In this case, assuming that every cross docking station can supply every geographic region, the possible paths throughout the network are sixteen. In reality, due to different constraints, every cross docking station can only supply a subset of the regions. The model handles this by assigning a very low bid to all those paths that are infeasible.

After bid calculation, the model selects the best bid and records the results. At the end of the simulation, all the orders have been assigned a route throughout the distribution network.