*1.1. AM as an Enhancer of Potential Cost Reductions in the Context of Companies with Large Inventories*

Companies providing supplies of spare parts are usually obliged to hold large sets of stocks. These inventory parts cause many issues to the companies, in the form of costs, warehouse space and logistic implications. Additive manufacturing (AM) technologies are perceived as a very powerful tool to address mass product customization, delocalized production and short series manufacturing, by maneuvering in relatively low levels of cost and short delivery periods for short series of products. This means that AM technologies could change the way that the companies supplying spare parts organize their business.

However, there is a lack of detailed studies on how the AM technologies impact the cost levels of inventories held by companies. The common case studies focus on the improvement and switch in production methods for very specific selected parts that, of course, yield major potential results. However in effect, all these studies lack of contextualization on the form of the relevance of those

case studies for the overall results in the company. Issues such as (i) the representativeness of the modified parts of the overall set of products, (ii) the cost impact of the part once sold in the overall of the company performance or (iii) the importance of the part studied in the business area of the company are usually not analyzed in detail. This means that the results cannot be extrapolated from the engineering case to the business case or, in the cases it be done, that normally the extrapolation is of relatively low economic impact.

Therefore, the present work addresses the whole matter, providing an engineering case study of a relevant part applied to a relevant business area. This article assesses the technical feasibility of implementing AM technologies in the manufacturing of spare parts. In particular, the assessment is applied to a case that it is representative of the taxonomy of a relevant company for spare parts providing in the fluid conduction systems sector. It undertakes the AM production process for the initial (conventional) design and evaluates the cost levels in comparison with the conventional manufacturing technologies. Furthermore, it performs a redesign of the case study part by means of topological optimization, validating the results both by means of computer aided and physical material testing methods.

Moreover, the study case contains the study of the context characteristics and consequences of the engineering case. It enunciates a frame for cost modelling that can be extended to other companies for further comparison and analysis, it analyzes the cost factors for the parts in the inventory and it finally produces an estimation of the possible impact of the utilization of AM over the cost levels of manufacturing and stocking of spare part products.

This work does not address the design of a new inventory policy in the sense of the number of units to be held. Instead, the present work analyses the possible cost reduction while maintaining the same quantities of stock parts. In this manner, the purpose of the present study is to quantify, the potential impact of the application of AM technologies to the supply inventory parts in a relevant context in a realistic manner (not optimistic or pessimistic). The significance of the work relates to the fact that this refers to a real engineering case study with relevant data, that could be used in future references for benchmarking the state of the art as the frame and methodology can be replicable with different products and technologies.

#### *1.2. Research Background and Motivations*

#### 1.2.1. Additive Manufacturing: Production and Design

AM technologies, also commonly referred as 3D printing technologies, comprise numerous different techniques concerned with the materialization of three-dimensional digital models as physical objects, without the need of molds or special tooling. All these techniques have a process in common—that the material is added and consolidated layer by layer, which is the reason why many authors refer to them as additive layered manufacturing technologies. This specific manner of construction consolidation yields a specific directional material behavior effect in the direction in which the layers are added (usually taken as 'z' direction, and so named as 'z-effect'), thus making the parts to adopt an anisotropic behavior being usually the 'z' direction the one yielding the least traction capabilities compared to the other directions 'x' and 'y'.

AM technologies can certainly be classified by two different criteria, i.e., depending on the physical solution in the systems for (i) material incorporation and for (ii) energy incorporation. Usual material incorporation procedures include feeding the raw material from a point (0d), from a line (1d), of an entire layer (2d) or in a bed (3d). Usual energy incorporation procedures include applying it to the raw material in a single point (0d), within a line (1d) or within a plane (2d). Very recent approaches in holographic patterning is opening new possibilities for the application in (3d) [1]. The classification in the AM systems is crucial as it determines the materials that can be processed, the characteristics of the parts that can be obtained and the cost levels of each specific technology.

Generally speaking, AM technologies make possible to manufacture functional parts with complex design geometries, offering the possibility to optimize the mechanical properties of a part while minimizing the weight and so the energy required in their manufacturing [2]. To this regard, the topological optimization techniques deal with the material distribution within a part domain, where material density can be increased or reduced therefore changing the rigidity of each specific domain [3], and even materializing a continuous density graduation.

One particular set of AM technologies that has widespread since the expiration of many patent protections (during 2005–2010) is the fused filament fabrication technologies (FFF). In FFF, the material is incorporated in the part from an extrusion head, being this an application of both (i) and (ii) as a point (0d). FFF processing has relatively low-cost levels of hardware and materials, as well as a huge list of possible materials to be processed [4]. During the last years, industry and academia have researched and developed solutions that cover industrial needs and that are capable to address market niches, and so its use has spread and the penetration in working environments has deepened. These facts position very well the FFF technologies to be utilized as means for production in short runs of delocalized production.

The mechanical properties of the parts manufactured via FFF depend on different parameters such as the internal structure, the orientation in the construction platform and the generation of printing paths [5]. Because of its nature, parts manufactured by FFF technologies have an intrinsic danger of delamination between construction levels, which mean that the construction direction must consider the different working modes that the part will have. Because of the so-called z-effect, introduced above, FFF technologies are far from achieving isotropic parts and its mechanical properties differ from the different construction planes [6]. In the literature, this has been extensively assessed and quantified, especially in computer aided engineering and under static loading conditions. Furthermore, some studies have been conducted to assess the behavior of printed parts under dynamical loads; yielding results more adjusted to the real physical performance of the parts [7].

Moreover, the transition of conventional manufacturing technologies to AM procedures imply many added challenges in the fields of design and materials use for each specific AM technology [8]. Again, there is a very important need for the topological optimization of the parts, which is key for decreasing the costs of the parts, the environmental impact as well as the material usage. This optimization process must be addressed during the design stages of the parts. It requires a detailed study of the working conditions considered for the part and it is normally undertaken by setting an objective function with the constraints of: Material properties, geometrical characteristics, design domain, loads, and supports. The optimization itself is performed by means of computing, iterating solutions between different models. However, it is necessary to allow some human design input to conduct certain aspects of optimization, which is the reason why it is so needed to have tools that allow to ease the free-form design [9]. The overall result of the design for additive manufacturing (DfAM) methods are three-dimensional models redesigned, normally with complex structures and capable to be manufactured by a specific AM technology.

Quantifying the impact of the possible substitution of a conventional manufacturing technology for an AM technology is an open technological challenge that needs to involve both technical knowledge (product optimization) and economic analysis (costing policies). The recent times have been of a high increase in the sales of desktop manufacturing systems—the units believed to be already sold in 2017 nearly double the figures of two years earlier, reaching over 528,952 units worldwide [10], so technologies such as FFF are now more available than ever to industrial companies. Now the challenge is to explore up to which point the current technology can reach relevant product materialization.

#### 1.2.2. Cost Modelling of Additive Manufacturing Technologies

The manufacturing costs levels of manufacturing AM technologies can be characterized relating it to three different factors: (i) Part weight, (ii) part dimensions, and (iii) construction time.

Part weight (i) is a common improvement claim for AM from conventional manufacturing technologies. This is because AM technologies only consolidate the material that would be a slice of the product. The material not used can be reutilized, thus saving raw material. However, the reality is that the material consumption to be accounted is higher in some AM methods. For example, in bed technologies, such as selective laser sintering (SLS) there are limits on the number of recirculation's of material in the bed, so further material consumption must be considered. Fruthermore, some deposition technologies, such as FFF, require printing supports (normally in the form of a honeycomb) to manufacture some of the parts, that will also be a scrap material rate.

Concerning (ii) and (iii), some industrialists prefer to treat them autonomously and some combine them in the elaboration of product's quotations. The reason is that the main factor for the cost in the volume of the part is the 'z' direction of the part when set in the machine for 3D printing. Moving the 3D printer head in the 'X'-'Y' plane can be very fast, but the maximum amount of time is spent when moving head and bed relatively over the 'Z' direction. Some authors have studied in detail the consolidation parameters, for example in selective laser melting [11], at all levels: Track, layer, and 3D object. The so-called 'hierarchical approach' facilitates the obtention of high-quality high-density parts. In this context, some industrialists prefer having construction time (iii) as a separate cost factor to integrate better the setting-up and post-production costs while others handle them in combination to have easier cost models.

With this rationale it is clear to see that, in AM technologies, complexity does not affect the final manufacturing cost of the part. Because of this, it is significant to undertake DfAM analysis before performing the manufacturing process, so to minimize the manufacturing cost levels.

In previous works of the authors [12], the costs configuration of the AM production technologies has been formulated in function of: Machinery costs, materials costs, energy consumption costs and labor costs. Implicit in these terms there are the factors of mass, 'z' dimension and construction times. Some other costs models [13] differentiate the well-structured direct production costs (labour, materials, equipment, etc.) from the ill-structured production costs (construction failures, transportation, inventory, etc.).

Having revised all this, in the present study, the Hopkinson and Dickens method [14] has been selected as the reference framework for manufacturing costs calculation. This method calculates the costs by splitting assumes that the energy consumption costs of the machines are negligible (assuming less than 1% of the final cost). Furthermore, the present work tries to broaden the work scope to the costs of manufacturing plus the costs caused by its stockage. Concerning this, some authors have prepared product lifecycle models for the spare parts sector [15], as it keeps being an active working topic.

#### 1.2.3. Cost Modelling and Issues of Holding Stock Parts

The evaluation of the inventory cost is a field widely addressed in the literature because of its many implications in the product supply chain. The economic order quantity (EOQ) is the more extended model used by the authors [16]. EOQ is a model that addresses how much product to order, taking into consideration the ordering costs and the holding costs. The ordering costs include some cost elements such as labor and other indirect office costs that enable to process the order. The holding cost includes the costs of storage, insurance, spoilage and others. The cost of capital is usually considered in the holding cost calculation [17] although some authors advocate for maintaining it as a separate cost factor.

Traditionally in the EOQ model, the holding costs are modeled in function of the average number of units per order. This term represents the stock cost [18,19]. Some authors introduce variations in this model, while others [20] consider two parts of a holding cost; i.e., one depending on the average number of units per order and another one depending of sudden increases in cost such as renting or renovating a warehouse for keeping extra units of product.

The inventory policies for controlling and maintaining optimal inventory levels study the balance between expenditure in ordering costs and in holding costs. This is the reason why it is necessary to develop and select the most effective inventory model yielding the optimum inventory framework minimizing its cost. Some authors [21] have assessed different models, such as lot for lot, EOQ, period order quantity (POQ), least unit cost, least total cost, least period cost, and the Wagner-Whitin model. After the assessment, this last model yielded the best results for the least total annual inventory cost.

In the frame of EOQ theory a cost model with arbitrary function is also developed [22]. In this case, the time-depending holding cost is introduced, in order to take into account the higher money effort for keeping fresh some perishable goods. Some other authors develop more complicated frameworks [23], incorporating the demand modelling as distribution functions [24], addressing the shortage implications in the ordering [25] and solving optimization problems using metaheuristics [18], genetic algorithms and multicriteria analysis [26].

Based on the literature review, there is a clear industrial need to be addressed in the present work, concerning the usual industrial parameters for decision taking; namely: Costs of labor, regional costs, and the time dimension. In the models reviewed, labor and regional costs are normally merged into the holding costs calculation. Therefore, there is a need to separate the accountability of such costs to compare the product costs when locating the supply chain in different geographical facilities.

#### 1.2.4. Case Studies and Implementation of AM Technologies to Industry Parts

Case study research (CSR) is a very suitable methodology for undertaking assessments in both the fields of engineering technology and production administration, which has been gaining importance over many different disciplines during the last years [27]. Compared to pure experimentation, which is probably the most common engineering approach, CSR can handle with the propositions on 'how' and 'why' of contemporary data without having to control behavioral events [28].

The study cases of application of AM technologies to product manufacturing in the recent literature have commonly addressed metal AM applications, and are characteristics of the companies that manufacture and sell AM equipment [29]. Some pure research experimentation approaches are capable of assessing in more detail the conditions and optimal parameters form parts obtention [30]. However, these very comprehensive studies focus on the technical product optimization that can be achieved, and do not assess whether the products improved have a defining impact on the company which produces them at its overall level. The reality is that, at the present time, most of the products optimized in such case studies are parts that are only product prototypes that are not in the main product core range of the firms. Alternatively, in some cases that the products are in the core activities of the companies, some parts fail to be economically relevant in the broader perspective.

The case study formulations can be prepared in single-case designs or multiple-case designs [28]. The results obtained in the engineering case are extrapolated and the impact is evaluated at the company level. Due to the nature of the case study, there has been a very comprehensive process to select the unit of analysis to make sure that the results can be extrapolated to the effect in the size of a company.

#### **2. Materials and Methods**

The presented case study methodology is grounded in the real circumstances of *Unistral Recambios*, a company that belongs to the *Fluidra* group, which accounted for a total price list cost of spare parts held in the inventory books over 15 M€ in 2017 (considered the year of the study). In such example, the large volume of references does not allow to treat all the cases one by one. In the present work, the decision taken has been to formulate a case study with a holistic (single unit of analysis) single-case design. The results obtained in the engineering case are extrapolated and the impact is evaluated at the company level.

Due to the nature of the case study, there has been a very comprehensive process to select the unit of analysis to make sure that the results can be extrapolated to the effect in the size of a company. This selection of the unit of analysis (segmentation and taxonomy) has been conducted in the phase of Study definition. In parallel to the study definition, there has been constructed the framework construction for costs evaluation.

The core of the study development has been performed in the engineering case, assessing a product relevant from the global set. The final phase is the analysis of the Impact of the optimization level achieved and the discussion of the possible extrapolation to the potential impact in both company and sector levels. From these, further achievable impact could be explored by the extension of the analysis to other product fractions and/or with the introduction of changes in the inventory policies of the company.

The overall flow diagram of the methodology undertaken in the study is presented in Figure 1.

**Figure 1.** Flow diagram of the present study methodology.
