Application of Cluster Analysis to Investigate 3D Printing Feasibility for Supplying Spare Parts in the Appliance Industry

INTRODUCTION

Based on computer-aided design (CAD) and computer-aided manufacturing (CAM) models, three-dimensional (3D) printing is a technology used to fabricate 3D objects by depositing successive layers of materials along digitalized toolpaths until a final structure is created. 3D printing includes a family of processing technologies originally developed to minimize lead times for prototyping with soft materials, such as plastics and wax, which are available in formats such as liquid polymers, solid powders, sheets, and thin wires. The physical prototypes produced through 3D printing enable engineers to test design ideas, fit options, and functionality (Berman, 2012). With advancements in material science and engineering, 3D printing is now widely used across various industries to produce functional end-use parts with strong materials, including metals and ceramics. According to Wohler Associates (2018), approximately one-third of 3D-printed parts were created for functional applications, with companies such as Airbus, Boeing, Honeywell, and General Motors (GM) exploring this potential. Overall market growth indicates that 3D printing technology could become a critical component of future supply chains.

For original equipment manufacturing (OEM) industries, supply chain management is essential and must be seamlessly integrated into the enterprise management roadmap to ensure the delivery and production of high-quality products, ultimately achieving business excellence. Unlike the regular procurement of large batches of materials and supplies for production, the OEM spare parts business supplies “spare parts” to customers to maintain and repair equipment. These spare parts are also known as “service parts, repair parts, or replacement parts.” This business has unique characteristics. First, because of the maintenance and repair focus, the primary goal is to provide quick replacement parts to customers to minimize equipment downtime. Secondly, because the need for spare parts is directly tied to equipment failure (rather than customer demand), it creates an uncertain, intermittent, and sporadic demand pattern in the service parts supply chain (Martin et al., 2010). Additionally, excess inventory is generally considered wasteful, and minimal inventory levels are preferred and practiced by manufacturing companies. These factors create significant challenges in managing stock levels to supply essential spare parts at minimal cost and with the shortest delivery time (Syntetos, Keyes, & Babai, 2009). Traditional supply chain practices of ordering and storing spare parts in batches often result in long lead times, lengthy customer wait times, and, consequently, dissatisfaction, alongside high inventory costs.

3D printing offers significant advantages for producing customized parts quickly and is particularly suited to small order quantities. This technology has been applied in various industries, including automotive, aerospace, food equipment, and medical and dental sectors, for a range of purposes. However, while 3D printing has existed for more than three decades, little effort has been directed toward understanding its potential impact on supply chain management within the appliance industry’s spare parts business.

LITERATURE REVIEW

Originally developed for mechanical design and manufacturing, 3D printing applications have expanded into a wide range of fields. Beyond parts design and prototyping, one of the earliest applications of 3D printing was in manufacturing tooling, such as complex thermoset molds, injection molds, casting cores, and inserts. These items can be 3D printed without concerns over rising costs or geometric limitations, as 3D printing often provides significant advantages over traditional machining methods. With the advent of open-source software in 2009, individuals began creating and printing objects with 3D technology, sparking exponential advancements (Marquardt & Zheng, n.d.). Now widely regarded as both an emerging and accessible technology, 3D printing has been extensively used in the medical field to produce human body parts, such as knee joints and dentures for implantation, because of its precision in design and customization at a reasonable cost. Additionally, 3D printing has been applied in bone tissue engineering, involving the fabrication of scaffolds that combine cells and materials to repair bone defects (Haleem et al., 2019). Studies also show that 3D printing has entered the construction industry, replacing manual labor and offering promising solutions for faster, more cost-effective, convenient, and environmentally friendly construction practices (Tabassum & Mir, 2023).

Being an essential component of smart and digital manufacturing, 3D printing could potentially bring disruptive innovations that may impact the logistics of global supply chain management. Not only is it able to provide customized products, it also shows advantages in supply chain management in terms of reduction in inventory, shipping costs, and capital spending on factories and warehouses. In the manufacturing and manufacturing service sectors, Sasson and Johnson (2016) noted that shifting from traditional supply chain methods to 3D printing may bring substantial benefits to organizations in flexibility, quick response to demand fluctuations, and on-location production. According to Garrett (2014), this type of transformation improves responsiveness to consumers and brings more agility to the supply chain. Mohr and Khan (2015) noted that 3D printing technology promotes a shift from offshore to domestic production and reduces dependence on suppliers overseas. Because of these characteristics, researchers considered 3D printing the most disruptive technology that will impact global supply chains. Waller and Fawcett (2013) recognized the four disruptive characteristics of 3D printing: quality, the need for less capital investment to launch a broad range of goods, a shift to a business model wherein consumers also become producers, and low economies of scale. Similarly, Nyman and Sarlin (2014) emphasized four different aspects of 3D printing, including sustainable and green operations, flexibility, small economies of scale, and change in the business model. Sasson and Johnson (2016) noted that production costs and lead times can be reduced by a hybrid supply chain model in which high-volume parts are produced with traditional mass production and low-volume, disruptive parts with the 3D printing method. These features will change how companies look at traditional supply chains.

An extensive literature review performed by Mohr and Khan (2015) captured seven different impact areas of 3D printing on supply chain management, noting these impacts as “mass customization, changing view on resources, decentralization of manufacturing, reducing complexity, rationalization of stock and logistics, changing value-adding activities and disruptive competition.” Attaran (2017) stated that being the most promising technology, 3D printing will become a core factor in the production process, affecting the road map of supply chain management. This technology transforms supply chains by consolidating manufacturing locations and logistic centers, bringing manufacturing closer to end users and therefore significantly reducing the transportation cost (Thomas & Gilbert, 2014; Khorram & Nonino, 2017). Depending on specific business natures, 3D printing can be used as an alternative or supplementary supply chain method to produce parts in small batches with a “made to order” concept that allows companies to mitigate carrying high inventory levels and reduce transportation costs, and, furthermore, from an overall perspective, supply chain complexity could be significantly reduced (Petrick & Simpson, 2013; Janssen et al., 2014).

Researchers have noted that 3D printing has a great potential to change the landscape of supply management for spare parts business in general, although 3D printing may bring varied impacts to specific industries (Holmström et al., 2010; Khajavi, Partanen, & Holmström, 2014; Liu et al., 2014; Muir & Haddud, 2018). For the spare parts business, because of its nature of being small order quantity and a large part variety, 3D printing is more suitable and profitable to supply parts for spare parts. Muir and Haddud (2018) used an online survey to collect data from 69 participants to understand the perceived impact of 3D manufacturing on the fulfillment of spare parts demand, inventory, and customer satisfaction in medical industry supply chains. Bradshaw, Bowyer, and Haufe (2010) performed a qualitative study and specified that 3D printing can be used to manufacture service and aftermarket accessory parts for laundry machines and small domestic appliances, such as mixers, blenders, and cameras. Khajavi and colleagues (2014) studied the utilization of 3D printing technology on the service parts of military war aircraft, and their results showed that expanding 3D printing in the spare parts supply chain was not feasible because of the high cost of 3D equipment and low speed of production. Liu and colleagues (2014) investigated 3D printing’s impact on the aircraft spare parts supply chain by developing three different models, including those that were conventional, centralized, and distributed with data gathered from the literature, with a primary focus on the impact of 3D printing on safety stock in the aircraft spare parts supply chain; however, this lacked real-cost data.

In particular, Boysan et al. (2022) reported a case study in which two models were built to estimate the total inventory operational cost of a spare parts business for an appliance manufacturing company, with one model in which spare parts are supplied in traditional manufacturing method and the other using 3D printing. Using real industry supply chain management data, this study compared the total inventory cost of the two scenarios; additionally, a hybrid cost model, in which some parts were supplied in traditional manufacturing and some in 3D printing, was hypothetically developed to maximize supply chain cost savings. Continuing the research path and using cluster analysis, this paper explores this subject in more depth to distinguish which parts are suitable for 3D printing other than traditional manufacturing in supply chain management.

RESEARCH METHOD

Pareto analysis

By using categorical parameters, namely as part family, part category, product platform, and part functionality, preliminary data analysis was conducted to examine the 49 “3D Printing preferred” parts and organize the groups through the Pareto chart. Figure 1 shows that 30 out of 49 parts are knobs, which correspond to 61% of the “3D printing preferred” parts. The top four part families, including knobs along clips, bushings, and covers, made up 76% of the “3D printing preferred” parts. The remaining part families have one part each.

Figure 1.

Categorization of part families for 3D printing preferred part.

In Figure 2, when the 49 parts were reviewed according to part category, 92% of the 3D printing preferred parts were from clothes washers, electric ranges, gas ranges, and refrigerators, while only very few parts were from the other three categories.

Figure 2.

Categorization of part category for 3D printing preferred parts.

Figure 3 shows the categorization of the 49 parts by appliance product platform; 51% of 3D printing preferred parts were from the food preparation platform. Fabric care was the next highest, with 31% in that platform. These two platforms covered 82% of the total population of parts. When the 49 parts were organized according to functionality, it showed that 96% of the “3D printing preferred” parts were nonfunctional (Figure 4). Functional parts are the parts that are directly involved in the appliance’s core function and behavior. The nonfunctional parts relate to the appliance’s quality attributes and constraints and how it performs its functions. An appliance might still function without optimal nonfunctional aspects. For example, for a refrigerator, the compressor, evaporator coils, and condenser are functional parts, while door seals, door handles, and shelves are nonfunctional.

Figure 3.

Categorization of parts by product platform (3D printing preferred = 49 parts).

Figure 4.

Categorization of parts by functionality (3D printing preferred = 49 parts).

Applying cluster analysis

Secondly, cluster data analysis was conducted using Minitab by including the four numerical data parameters—namely, dimension, weight, EAU, and MOQ—with no predefined groups to decide the potential number of clusters. Figure 5 shows the results of the cluster observations function. The amalgamation steps in cluster analysis show the similarity level of the 3D parts’ numerical data decreases and the consecutive distance between the joined clusters increases as new clusters join.

Figure 5.

Clusters observations for numerical data values (3D printing preferred = 49 parts).

Initially, the similarity level decreases by increments of less than one (from 99.9898 to 99.0493) until step 37; the similarity level decreases by increments of less than two (from 99.9898 to 98.0381) from step 37 to step 41. However, the similarity level changes significantly after step 43 compared with previous steps. For example, similarity decreases by more than seven (from 94.6539 to 86.9703) from step 45 to step 46, by 30 from step 46 to step 47, and by even more than 60 (from 86.9703 to 56.5340) from step 47 to step 48. The marginal change in similarity level from steps 1 to 45 suggests that 44 of the 49 “3D printing preferred” parts are highly similar and can be grouped into a single cluster. The distance and similarity changes across steps indicate that four clusters are potentially enough to categorize the numerical data values for the 49 “3D printing preferred” parts.

Based on the cluster observations, the amalgamation steps displayed in Figure 5, and using the 94.6% as similarity cutoff value, the output of the cluster analysis from Minitab with four clusters is displayed in Table 3, including the final partition, cluster centroids, and distances between cluster centroids. The distance from centroid, which does not have a physical unit, is a mathematical measure (Euclid distance) of how different the clusters are based on the data metric used in this study. The smaller values mean that clusters are very similar, and higher values mean they are more distinct. Figure 5 displays the groups in a tree diagram formed by clustering observations at a similarity level of 94.6%.

Table 3.

Results of Cluster Analysis (3D Printing Preferred = 49 Parts)

Final Partition
	Number of Observations	Within Cluster Sum of Squares	Average Distance From Centroid	Maximum Distance From Centroid
Cluster 1	44	4,014,088	212.217	923.597
Cluster 2	1	0	0.000	0.000
Cluster 3	3	503,321	342.400	502.208
Cluster 4	1	0	0.000	0.000
Cluster centroids
Variable	Cluster 1	Cluster 2	Cluster 3	Cluster 4	Grand Centroid
Weight (ounce)	2.273	1.4	0.02	0.3	2.08
Dimension (cubic inch)	26.386	3.7	0.32	42.9	24.67
EAU	34.19	32.4	26.40	20.6	33.40
MOQ	277.227	23000.0	2500.00	10000.0	1075.47
Distance Between Cluster Centroids
	Cluster 1	Cluster 2	Cluster 3	Cluster 4
Cluster 1	0.0	22,722.8	2,222.9	9,722.8
Cluster 2	22,722.8	0.0	20,500.0	13,000.1
Cluster 3	2,222.9	20,500.0	0.0	7,500.1
Cluster 4	9,722.8	13,000.1	7,500.1	0.0

The 49 “3D printing preferred” parts are classified into four clusters based on cluster analysis. Table 4 summarizes the cluster analysis result, according to the count of part number, average weight, average dimension (volume), average estimated annual usage (EAU), and average monthly order quantity (MOQ). Cluster 2 and cluster 4 contained only one part each; cluster 3 contains three parts: PN11, PN14, and PN29. The first cluster included 44 parts with an average weight of 2.27 ounces, an average volume of 26.39 cubic inches, an average EAU of 34 pieces, and an average MOQ of 277 pieces. Cluster 2 and cluster 3 showed a lower average weight, dimension, and EAU but higher MOQs compared with cluster 1. Cluster 2, cluster 3, and cluster 4 all showed significantly high average MOQ values (see Table 3). The cluster analysis showed that perhaps because of MOQs that were high but different in weight, volume, and EAU, the five parts were classified into three different clusters. In general, for each cluster, the average MOQ is much higher than the average EAU, indicating that a lot of spare parts need to be held in the inventory. If 3D printing were used for supplying these parts based on instant customer orders, then fewer spare parts would be sitting in the warehouse, thus reducing inventory costs.

Table 4.

Summary of Cluster Analysis (3D Printing Preferred = 49 Parts)

Clusters	Count of Part Number	Average Weight (Ounces)	Average Volume (Cubic Inch)	Average EAU	Average MOQ
1	44	2.27	26.39	34	277
2	1	1.44	3.75	32	23,000
3	3	0.02	0.32	26	2,500
4	1	0.32	42.92	21	10,000

Furthermore, Table 5 shows another way of possible grouping of 49 parts based on four numerical data parameters, including weight, volume (dimension), EAU, and MOQ. In Table 5, 67% of the “3D printing preferred” parts (33 parts) had fewer than 1,000 pieces MOQ, fewer than 100 pieces EAU, and less than a 50-cubic-inch volume and weighed less than four ounces.

Table 5.

Grouping Parts by Numerical Parameters (3D Printing Preferred = 49 Parts)

Weight Classification (Ounce)	Volume Classification (Cubic Inch)	EAU Classification	MOQ Classification	Part Count	Percentage
0–4	0–50	0–100	0–1,000	33	67%
			1,000–3,000	4	8%
			>10,000	2	4%
		100–200	0–1,000	3	6%
	50–150	0–100	0–1,000	2	4%
4–11	0–50	0–100	0–1,000	3	6%
	50–150	100–200	0–1,000	1	2%
	150–350	100–200	0–1,000	1	2%

Grouping of “3D printing preferred” parts

Four clusters are suggested to classify the “3D printing preferred” parts by implementing cluster analysis with all data without distinguishing the numeral parameters defined in the study. Next, the 49 “3D printing preferred” parts are grouped based on the numerical parameter to determine commonalities. Figure 7 shows the grouping of the 49 parts in terms of part family, part category, functionality, part platform, volume, weight, MOQ, and EAU. The commonalities observed are as following:

• •

  The vast majority of “3D printing preferred” parts have smaller volume and weight.

• •

  Their EAU is between 0 and 100.

• •

  Their MOQ is less than 1000.

• •

  The vast majority of them are nonfunctional parts.

• •

  Food preparation, food preservation, and fabric care are the major production platforms.

• •

  Knobs are the major part family.

• •

  Clothes washers, electric ranges, gas ranges, and refrigerators are the major part categories.

Figure 6.

Dendrogram of cluster analysis results (4 clusters).

Figure 7.

Grouping of 49 parts (3D printing preferred).

CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE STUDY

This research explores the potential of using 3D printing technology to substitute traditional manufacturing batch production for supplying spare parts in the appliance industry. Using the spare part data from the appliance industry, this study investigated the provision of spare parts in the spare parts division of an appliance manufacturing company.

The first stage of this study built two inventory cost models that compared the current supply chain management with the traditional batch manufacturing strategy and the hypothetical supply chain management model with the 3D printing technology. Comparisons of the two models identified 49 parts out of 100 sample parts that were 3D printing preferred because they showed an advantage of cost savings with the 3D printing technology. Continuing the first stage of the study, this research further applied cluster analysis and Pareto analysis to build groups based on the commonalities or similarities among the “3D printing preferred” parts. The cluster analysis suggested four clusters can be considered in terms of similarity level, grouped according to numeral parameters (volume, weight, EAU, and MOQ). The analyses found that the vast majority of the “3D printing preferred” parts are nonfunctional, in small dimensions and weight, and with a small annual usage number and a small monthly order quantity. The research also found that for many “3D printing preferred” parts, their MOQ is significantly larger than the average annual usage level. That implies that many spare parts currently sit in the inventory warehouse for over a year and are a big waste. If 3D printing were used for supplying these spare parts, it would reduce the inventory costs.

This study revealed common factors of 3D printing preferred parts. These common factors included family, functionality, category, weight, volume, estimated annual usage, and minimum order quantities. Because this research was designed to identify parts with the highest supply chain potential in terms of cost, observed commonalities and similarities could be helpful to focus on a product mix in future analysis. Another significance is that this study demonstrated an example to understand the potential impact of 3D printing on supply chain management in the appliance industry and to make strategic decisions to determine which types of parts are more likely to be supplied by the 3D printing technology.

CHALLENGES OF IMPLEMENTING 3D PRINTING TECHNOLOGY FOR THE SPARE PARTS BUSINESS

Currently there are no existing guidelines from the appliances’ spare parts manufacturers regarding 3D printing and supply chain management. Therefore, the methodology and processes defined during this research could be used as a guideline to implement 3D printing for the spare parts business of the company. Secondly, part selection, data collection, data mining, and uploading data to the software platform in the correct format requires time, skills, and attention to detail. Digitizing parts and performing a technical feasibility analysis are required to determine whether parts are 3D printable, which takes time. Some parts require special surface finishes when a secondary process is needed to achieve the required quality specifications. In addition, the quotation process was challenging and lengthy because different types of materials and part families were involved in this research.

Besides the above challenges, it should be noted that the purchase cost of a single part is significantly higher using 3D printing compared with traditional manufacturing. In this research, only 49 out of 100 parts showed cost savings increase if the company switched from traditional manufacturing to 3D printing. The main reason is that the application of 3D printing in the spare parts business is still evolving, and several cost factors affect it negatively, such as material cost, utilization of technology, and technical capabilities. With technological advances, 3D printing may change the landscape of the spare parts business more favorably. Future research may focus on using a 3D printing platform to shape spare parts directly from a 3D printing supplier to customers. This is called decentralizing the spare parts supply chain, and there is a lack of research in this area in terms of total cost in a decentralized supply chain, with 3D printing playing a part in different industries.