In this chapter, the used methodology and data are to be explained.
3.1. Methodology
Within the tackled literature, various studies used different methods to reach their desired results. In a study about connecting CE and I4.0 [
57], the cause-and-effect relationship was used between various dimensions of I4.0 and CE in the supply chain area. The dimensions of I4.0 were obtained through the analysis of exploratory factors. A total of 161 responses from Indian manufacturing companies were the sample data. Additionally, they performed a cause-and-effect relationship through DEMATEL analysis. Other studies also addressed different hypotheses to be analyzed in a specific way. In a study that examined the role of I4.0 on CE practices and the capability of the supply chain to increase the company’s performance [
58], eight hypotheses were presented while structural equation modeling was used for analyzing them. Additionally, in investigating how I4.0 technologies and stakeholder pressure influence circular product design and impact company performance [
59], five hypotheses were assumed. Partial least squares path modeling for data analysis was used. In another study [
60], a qualitative analysis of selected case studies aimed to answer three research questions. The results were visualized to highlight applying digital technologies’ effects on processes, companies, products, and supply chain within the transition into CE. Likewise, in a study regarding adopting the I4.0 technologies pattern in manufacturing companies [
2], four hypotheses were presented about using smart manufacturing technologies. It included data analysis of a questionnaire survey of companies, so relevant to our research. The first step to analyzing the data was by identifying the tackled companies into several maturity scales regarding their adoption level of smart manufacturing technologies. Two groups with distinct technological levels were needed at a minimum for testing the hypotheses and finding out different patterns between these groups in order to explain the I4.0 adoption. Then, a hierarchical cluster analysis was used to determine the adequate number of groups for sample division. After having obtained the cluster compositions, an analysis of the demographic aspect of the cluster members was performed. Pearson’s Chi-squared test was used to reject the null hypothesis that stated that there is no association among the variables. Additionally, the test of Fisher’s exact was used for associations to reach four observations or fewer.
As we could find, there is a wide spectrum of approaches and methods used in the literature to investigate I4.0’s impact on CE. The rationale of our methodology is based on our research questions and available data. The general research question is whether there are relationships between the use of I4.0 and CE in manufacturing companies. The data comprise a sample of central European manufacturing companies, which includes the use of selected I4.0 and CE technologies. To find out if there are relationships, the raw data were filtered at the beginning to exclude any invalid entries, and then, we built our methodology in two steps. First step: grouping the data to see if there are some differences in the use of technologies in the subsample groups. For this, contingent tables were used. This helped to find out where to expect possible relations between technologies. In this step, we also included non-I4.0 technologies to see if there are differences between the use of I4.0 and non-I4.0 technologies. Second step: logistic regression (by IBM SPSS Statistics 25 software, Armonk, NY, USA) was used to validate the expected relations. Before starting the logistic regression, a correlation test was applied to the independent variables to affirm their independence. It should be mentioned that even in a case where logistic regression shows a statistically significant relation between I4.0 and CE technology, it does not reflect a causal relationship. Therefore, the odd ratio was used to reflect the strength of these possible relationships, but it cannot affirm them as a direct influence. In other words, with our methodology, we can only show relations, but not affirm if I4.0 supports or enhances the CE. This is one limitation of the used methodology. Nevertheless, showing the existence of the significant relationship can help other researchers to focus on this relationship and investigate causality.
3.3. Results and Discussion
This section is divided into two sub-sections for presenting the results of the two research models, 1 and 2.
3.3.1. Relations between the Use of I4.0 and CE Technologies (Research Model 1)
Within this model, possible relations for the impact of used technologies in the areas of production control, digital factory, automation and robotics, and AM technologies on the adoption of REW (resp. REE) technologies were analyzed. The differences between the companies’ percentages that use I4.0 (but also non-I4.0) technologies in the whole sample compared to the subsample of companies that are using REW (resp. REE) technologies in the manufacturing companies are presented in
Table 1 and
Figure 4.
By comparing the percentages for the whole sample and subsample of companies that use REW, we can find the highest differences in the case of three technologies (industrial robots for handling processes, near real-time production control system, and software for production planning and scheduling). There are also another three technologies showing differences (industrial robots for manufacturing processes, the digital exchange of product/process data with suppliers/customers, and systems for automation and management of internal logistics). Based on this, we expect relationships between the use of these technologies and the use of REW.
By comparing the percentages for the whole sample and subsample of companies that use REE, we can find the highest differences also in the case of three technologies (industrial robots for handling processes, near real-time production control system, and software for production planning and scheduling). There are also another three technologies showing differences (systems for automation and management of internal logistics, the digital exchange of product/process data with suppliers/customers, and mobile/wireless devices for controlling facilities and machinery. Based on this, we expect relationships between the use of these technologies and the use of REE.
The statistical test is applied to validate the expected relationship and support or deny the hypothesis. The method for testing is the logistic regression by IBM SPSS Statistics 25 software. For the H1a test, the sample was N = 543 after filtering the raw data. Before testing, a correlation test was applied to the 12 independent variables (see
Figure 1, I4.0, non-I4.0). The tackled variables (technologies) appeared to be independent where the highest correlation value was 0.3638 except for 3D1 and 3D2 technologies, which showed 0.533. These values allow us to consider the 12 technologies as independent variables. The results of the logistic regression are presented in
Table 2.
As we can see in
Table 2, four technologies of SPP, NRP, IR1, and IR2 showed statistically significant relationships with the dependent variable REW. IR2 and NRP showed the strongest significance of relationship and influence (Exp(B)) on the REW. Based on this, we can conclude that the significance of the relationship between the use of specific technology and the use of REW is not dominantly influenced by whether it is I4.0 technology or not.
To validate the expected relationship in H1b, we used the logistics regression test again. Before this analysis, we filtered the raw data accordingly (final N = 546 companies). The correlation test is the same as the previous one (same 12 technologies). The results of the logistic regression test are presented in
Table 3.
As we can see in
Table 3, four technologies of SPP, NRP, SAM, and IR2 showed statistically significant relationships with the dependent variable REE. IR2 and SPP showed the strongest significance of relationship and influence (Exp(B)) on the REE. We can conclude also in the case of REE (similarly to REW) that the significance of the relationship between the use of specific technology and the use of REE is not dominantly influenced by whether it is I4.0 technology or not. It is important to highlight that NRP showed a 0.106 significance result in
Table 3. It is even more than the significant step of 0.1, it is very close to it, therefore, it was considered equal to 0.1.
3.3.2. Relations between Use of I4.0, Non-I4.0, CE Technologies, and Product Characteristics (Research Model 2)
Within this model, possible relations for the impact of used technologies in the areas of production control, digital factory, automation and robotics, and additive manufacturing technologies on the new or improved product development (NPI), (resp. improved environmental impact (IEI) of the product) were analyzed. The differences between the companies’ percentages that use I4.0, non-I4.0, and CE technologies in the whole sample compared to the subsample of companies that have performed NPI, (resp. IEI) are presented in
Table 4 and
Figure 5.
By comparing the percentages for the whole sample and subsample of companies that have performed NPI, we can find that the highest differences are in the case of three technologies (NRP, MW, and VRS). Other technologies showed less significant differences, but interestingly, some were negative (the highest negative difference was SPP), but as value it was small. Based on this, we expect a relationship between the use of these technologies and the execution of NPI.
By comparing percentages for the whole sample and subsample of companies that have carried out IEI, we can find the highest differences in the case of five technologies (DS, SPP, PLM, VRS, and IR2). Additionally, two other technologies (IR1 and 3D1) showed moderate differences. Based on this, we expect more relationships between the use of these technologies and the execution of IEI.
To validate the expected relationship in H2a, we used the logistics regression test again. Before this analysis, we filtered the raw data accordingly (final N = 535 companies) and made the correlation test for 14 technologies (12 technologies + 2 CE technologies) (see research model 2 and
Figure 1. The highest correlation value for the two new variables (technologies) was 0.346, which is still low and allows us to consider the 14 technologies as independent variables. The results of the logistic regression are presented in
Table 5.
As we can see in
Table 5, only the two technologies of VRS and 3D1 showed statistically significant relationships with the dependent variable—NPI. They both showed strong significance of relationship and influence (Exp(B)). Based on this, we can conclude that surprisingly, the significant relationships are not between technologies that we expect according to the differences identified above (
Figure 5), but the main finding is that it seems that I4.0 technologies dominate over non-I4.0 and CE technologies in having significant relationships with the execution of NPI in manufacturing companies.
In the last part of the analysis, to validate the expected relationship in H2b, we used the logistics regression test again. Before the analysis, the raw data were filtered accordingly (final N = 430 companies). The correlation test is the same as the previous one (same 14 technologies). The results of the logistic regression are presented in
Table 6.
As we can see in
Table 6, four technologies of SPP, PLM, VRS, and IR2 show statistically significant relationships with the dependent variable—IEI. The strongest significance of relationship and influence (Exp(B)) on the IEI has PLM. Interestingly, despite the expected higher number of technologies to be related to the execution of IEI, the regression does not prove it. Nevertheless, in contrast to the execution of NPI, it seems that the significance of the relationship between the use of specific technology and the execution of IEI is not dominantly influenced by whether it is I4.0 technology or not.
3.4. Discussion of the Results
The investigation of the relations between the use of I4.0 and CE technologies (research model 1) showed that in general, it seems that both I4.0 technologies and non-I4.0 technologies could have significant relations with CE technologies (in our study REW and REE). Interestingly, both have significant relation with three identical technologies (IR2, NRP, and SPP) and one different for each. The most significant relation (measured by Sig. and Exp(B)) (
Table 3 and
Table 4) in the case of both CE technologies is IR2, i.e., industrial robots for handling processes. This relation could be possibly connected to the technological level of the company. The existence of the relation with the second identical technology (NRP), for both CE technologies (especially the REW) could be caused by specific characteristics of the production process. The third commonly related technology (SPP) (especially significant for REE) can support previous arguments, that the company that uses REW or REE should be on some technological level and have a specific production process, where it can apply SPP. In the case of REW, there is one different significant technology (IR1). Explanation of significant relation with IR1 in the case of REW can lead us to the sectors such as automotive, electronics, etc., where the use of IR1 is widespread, so again to some specifics of the production process. In the case of REE, the different technology is SAM. We can only assume that some specifics of production process can take a role in this relation.
The results show significant relations of CE technologies (REW and REE) with robotics (IR1 and IR2), which is in partial agreement with e.g., the review of [
7], who stated that there is most evidence of the positive impact of additive manufacturing and robotics on circularity in companies. This found relation (in the case of IR2) is also in accordance with Álvarez-de-los-Mozos et al. [
48] and Renteria et al. [
49]. However, another study [
4] stated that additive manufacturing could be exploited to improve energy consumption, which is not in line with our results. Another finding [
7] that showed AM and VRS having the potential to reduce energy consumption is also not supported by our results. In addition, they showed for robotics that CE energy indicators vary in a range between 1.7 and 2.7 (on Likert-scale 0–4), i.e., the value of the influence is medium-high, however, the impact on the CE water variable has been less valued than 1.7. Our results indicate the opposite situation since in our case REE has a significant relation with only one robotics variable (IR2) and REW has a significant relationship with both robotics variables (IR1, IR2), but we should be aware of the different methodologies and variables in both studies. Nevertheless, our results are in line with additional findings of [
7] that identified small energy reductions (less than 5%) in relation to the use of robots, despite the energy consumption of the robots. When looking solely at the REE technologies, a significant relationship is found with SPP and SAM, in accordance with Rosa et al. [
4], Bloomfield et al. [
5], Lahrour et al. [
28], and Leino et al. [
29] and in case of NRP with Rosa et al. [
4] and Hatzivasilis et al. [
33]. When looking separately at REW technologies, significant relation was found with NRP that is in line with Rosa et al. [
4] and Hatzivasilis et al. [
33], and in the case of SPP with Rosa et al. [
4], Bloomfield et al. [
5], and Nascimento et al. [
30].
The investigation of the relation between I4.0, non-I4.0, and CE technologies and execution of new or improved product development (NPI) (resp. new or improved products with improved environmental impact (IEI)) (research model 2) showed major differences. In the case of NPI as dependent variable, two technologies (VRS and 3D1) showed statistically significant relationships. Here, the explanation is quite clear since both VRS and 3D1 are logically tight to the product development. Moreover, this result confirms the validity of our data and analyses. Lastly, it should be mentioned that clear dominance of I4.0 technologies appears here. In the case of IEI as a dependent variable, the situation is different. Similar to NPI, VRS (as a product development tool) created significant relations with IEI. Nevertheless, even higher significance (also influence (Exp(B)) is in the PLM (Product lifecycle management or product/process data management). These are important findings, that PLM has the potential to be an influential factor in the improvement of the environmental impact of the products. There were also another two technologies (SPP and IR2) that showed statistically significant relationships with IEI. This is not so straightforward to explain, but we assume similarly to above, that it can be connected to the technological level of the company and specifics of the production process. Lastly, it should be mentioned that no clear dominance of I4.0 technologies over non-I4.0 was found. In addition, it seems that there was not a clear connection between the use of CE technologies and the development of a product with improved environmental impact.
Our results in the case of NPI (as a dependent variable) support the literature findings [
2], which showed that there is a connection between the adoption of Smart Manufacturing and Smart Product technologies. Our finding on VRS relation to product development was also in accordance with Rosa et al. [
4], Kuik et al. [
36], and Wang et al. [
37]). Another study [
4] showed that I4.0 technologies can have a positive effect on the lifecycle management of products, while we found similarly that the use of virtual reality and robotics is related to the development of the IEI (improved environmental impact of a new product) by the company. The results [
7] that show robotics to have a medium influence (1.5–2.2 on Likert-scale 0–4) on reuse, and recovery characteristics of the products are also in agreement with ours since we identified the relationship between the use of robots (IR2) and IEI. Moreover, the relationship between IEI of the product and VRS technology is in line with the findings of Kuik et al. [
36] and Wang et al. [
37], while the relation with IR2 is in accordance with Álvarez-de-los-Mozos et al. [
48] and Daneshmand et al. [
50]. Finally, a relationship was found between IEI with PLM and SPP is also supportive of previous studies (Rosa et al. [
4] and Unruh [
27]).
The results of our analysis, from the view of tackled I4.0 and non-I4.0 technologies, showed a statistically significant relationship with dependent variables (REW, REE, NPI, IEI) in a few of them. Interestingly there were only two technologies (SPP and IR2) that showed a significant relationship (so potential impact) on the CE technologies (REW, REE) but also on the development of the product with improved environmental impact (IEI). What is behind this wider “pro-environmental” scope of these two technologies (in comparison to others) is not clear but could guide the focus of future research in this field.
Regarding the validity and limitations of this study, three issues should be considered. First, the used data were collected from the manufacturing companies only, which excludes other areas where applied I4.0 technologies can have an impact on CE such as service, health, transportation, and education sectors. Second, despite the questions used in the survey were formed in a few steps and pre-tested, invalid answers or human mistakes can happen. Third, the data were collected in 2018. While only four years were passed on this survey, fast development in this topic is expected.