Design for Additive Manufacturing: Cost Evaluations

As the recent literature reviews by Kadir et al. [5] and Costabile et al. [6] show, previous studies use a wide variety of models and techniques to estimate unit costs of products when adopting additive manufacturing. Some of them limit the analyses to the three main cost drivers affecting the total cost, namely machine, labour, and material costs, while others also include additional drivers, such as logistics, tooling, electricity, set-up, and/or inventory costs [5, 7, 8]. The approach adopted for the cost calculation too varies among the studies. Method-based, task-based, and level-based techniques are employed in the literature, according to the perspective adopted for cost classification, namely, finance, manufacturing, or management [5]. Finally, as highlighted by Thomas [9], the goal of cost calculation in previous literature is either the identification of the resources consumed in the process or the comparison with costs associated with traditional manufacturing.

In line with other studies comparing total costs in traditional and additive manufacturing environments, the authors adopted a task-based approach using a process-oriented cost model [10-13]. Following Lindermann et al. [14], they divided the production process into two main phases, representing two different cost centers: (1) pre-processing, which consists of all the activities needed as a prerequisite for effective production, and (2) part manufacturing, which deals with the production process itself. In selecting the cost drivers associated with each phase, the authors considered only the factors directly affecting the part cost, namely, those differing between traditional and additive manufacturing environments [10, 11]. The overall goal of the analysis was indeed not the calculation of the effective total cost, but rather the comparison between the two manufacturing environments. The costs included in the estimations were therefore six: CAM software preparation costs for the pre-processing phase and material, labour, machines, tools, and electricity costs for the part manufacturing phase. All the other costs, which do not vary among additive and traditional manufacturing environments, such as logistics, administrative overhead, rental costs, etc. [15, 16], were not considered in the model. This choice is common in previous literature, considering not only that their estimation would be an unnecessary complication but also that their effect on the total cost is typically limited to 10% [16, 17].

The analyses were based on some basic assumptions. First, the authors considered the total number of working hours per year equal to 3,840 [h/year], which results from a working time of 16 hours per day, 5 days per week, and 48 weeks per year. Second, they relied on a medium-sized Italian company (Company Alfa) operating in the engineering sector for all the estimations, concerning both cost drivers (e.g., tools, material, electricity, labour costs, etc.) and time requirements (e.g., set-up time). Finally, they calculated the machine cost using the straight-line depreciation technique [10, 11, 15]. This required the estimation of the total cost and the economic life of production machines. Dividing the total cost by the number of useful life hours and the machine cost per hour, which was taken into consideration for both set-up and production activities, could be derived.

The following paragraphs explain how the costs were estimated in the two manufacturing environments.

2.1 Cost estimation model for traditional manufacturing

In traditional manufacturing, the core production is carried out in two steps. First, in the roughing step, an initial rough block volume of steel is processed by subtracting the waste material to create the rough shape of the product. Then, in the surface finishing step, a further layer of material is removed to obtain the final shape of the product and meet the quality requirements. Since both steps are carried out on a CNC machine, two ad-hoc CAM programs for roughing and surface finishing must be developed in the pre-processing phase.

The procedure used to estimate the six cost drivers in traditional manufacturing (i.e., CAM programming, material, labour, tool, machine, and energy costs) is summarized in Table 1.

Table 1. Procedure for cost calculation in traditional manufacturing

The results described in the previous sections provide several cues for some interesting observations concerning the convenience of adopting additive manufacturing solutions.

First of all, by comparing Figures 1-3, it emerges that the cost of CAM programming has a very strong impact on the total unit cost when traditional manufacturing is adopted: it reaches the values of 77%, 48%, and 84% of the total unit cost for automotive gear, bottle blowing mould, and hip prosthesis, respectively, while in additive manufacturing, these percentages are significantly lower. As expected, when numerous batches are considered, the CAM programming cost can be spread among several units and becomes almost irrelevant to the total unit cost. It seems therefore that a bigger batch reduces the unit product cost in traditional manufacturing more than what happens in additive manufacturing. Not by chance, traditional manufacturing becomes more and more convenient when numerous batches are considered.

However, in general, it is not possible to state that the convenience of adopting traditional vs. additive manufacturing strictly depends on the number of pieces to be produced. The analyses suggest that the size and shape of the products influence the evaluation as well. As already highlighted in the reference [2], additive manufacturing is a rather slow manufacturing solution; as the volume increases, the manufacturing times increase as well, and so do the related machine costs. This is mainly the reason why three different situations emerge from the three product cases.

As far as the results of the hip prosthesis are concerned, the cheapest and quickest solution is offered by the use of additive manufacturing, especially if the geometry optimization is exploited. The product volume of the hip prosthesis is indeed very limited, and, for this reason, material and SLM machine costs are not particularly high. On the contrary, the CNC programming cost in traditional manufacturing is quite considerable, especially if compared with the other cost items, making this environment less convenient from an economic viewpoint. To further explore these results, the authors made some sensitivity analyses by changing the batch size, as they did for the other product types. Surprisingly, the results showed that additive manufacturing (with optimized geometry) represents the most convenient environment with a batch of 1,000 pieces, which is contrary to what happens with the other products. Obviously, the use of such a numerous batch for a hip prosthesis makes no sense from a practical viewpoint, but it corroborates the aforementioned observations: the limited product volume and relatively quick 3D printing time can overcome the limitations of the long CNC programming time needed for traditional manufacturing.

Considering the bottle blowing mould, the additive manufacturing represents a much more expensive and slower solution than traditional manufacturing. This is mainly due to the very large volume of the product, which significantly increases the manufacturing time required by the SLM machine and, consequently, the related cost. This latter, which accounts for almost 87% of the total unit product cost, exceeds the programming costs of traditional manufacturing, making additive manufacturing less convenient. The cost and time differences between the two solutions become even more significant as the production volume increases.

Finally, as regards the automotive gear, there is an intermediate situation. Both CNC programming costs and SLM machine costs are quite high, and even if traditional manufacturing represents the most convenient solution, the cost difference with additive manufacturing is rather limited for a single batch. However, when a batch of 1,000 pieces is hypothesized, the cost of CNC programming significantly drops, while the SLM machine cost does not change, making traditional manufacturing extremely convenient.