BUS606: Operations and Supply Chain Management

According to Sanchez, experimental design employs factors, which are independent variables, to test the impact on the outputs of the experiments, which are the response variables. As stated by Montgomery, this technique is useful to verify if a given factor influences, or not, the response of a system.

For Gomes and Costa, an experimental problem can be supported by the design of experiments and the statistical analysis of the data. There are several types of data collection strategies for DOE, and one of them is the complete factorial design.

Table 5 shows the control factors used in this paper.

Table 5 Control factors used in the proposed method.

Factor	Description	Level	Value
Features per part	Quantity of features in the part	Low	1 to 3 features per part
		High	7 to 10 features per part
Duplicated Features	Quantity of duplications of the same feature in each part	Low	1 to 3 features
		High	7 to 10 features
Batch Size	Quantity of parts in each batch	Low	10 to 50 parts
		High	200 to 500 parts
Type of Demand	Type of demand which will be applied to the system	Low	Random
		High	Controlled
Type of Plan	Process plan used in simulation (with or without alternatives)	Low	Plan With alternatives
		High	Plan Without alternatives
Simulated Batches	Total quantity of simulated batches in each replication	Low	50 batches
		High	200 batches
Arrival time	Arrival time from the current batch to the next batch	Low	5% to 10% of processing time
		High	20% to 30% of processing time
Maintenance	Maintenance concepts applied in the machines	Low	With maintenance
		High	Without maintenanc

The response variables that are considered: throughput (parts/hour), work in process (parts), resource utilization (%), final time (hours), total waiting time (hours) and lead time (hours).

In order to perform the analysis, three types of layout are considered: the job shop layout, the traditional cell layout and the virtual cell layout. Both process plans with alternatives as well as conventional plans (without alternatives) are considered. In the case of selecting an alternative operation, a penalty of 40% in the machining time is applied. An explanation for this penalty is that, if a hole needs to be made in a part, it may be performed in a milling machine or a drilling machine. Usually a milling machine will be chosen due to its versatility when compared with the drilling machine. But since the drilling machine can also be used to make the hole, it is included as an alternative in the process plan, but a less desirable one.

In order to reduce errors inherent to the variability in the execution of the simulation, 10 replications were performed for each combination of factors, that is, each experiment was replicated 10 times. Also, a full factorial design was used, meaning that all possible combinations were performed.

According to table 2 , there are eight control factors and, since only two levels were chosen in each factor (high and low), 2^8 = 256 experiments were tested. Since for each experiment ten simulations were run, a total of 2,560 experiments were performed. This set of 2,560 experiments was applied to each of the three different layouts (job shop, traditional cell and virtual cell), totaling 7,680 experiments. The total simulation times were one of the control factors of the experiment, and the warm-up time was 1h. That is, the first hour of each simulation was excluded from the statistical analysis.