Introduction

The UK National Strategy 2018–2025 for Additive Manufacturing (AM) estimates that by 2025, 60,000 jobs will be supported by AM and associated economic knowledge (Additive Manufacturing UK, 2017). Yet, one of the key barriers to the industrial progression of additive manufacturing (AM) is fragmented knowledge of design for manufacturability across AM platforms, parameters and materials. Knowledge transfer of AM is predominately bottom-up, leading to not only an inefficient, but an application-specific knowledge base (Thomas-Seale et al., 2018). Whilst knowledge transfer between all the facets, is a well acknowledged requirement of efficient design (Kusiak, 1993), there is an inherent deficiency in the breadth and/or depth of design for AM (DfAM) knowledge held by design engineers.

Innovative and efficient DfAM is equally dependent on creative freedom, underpinned by knowledge of the manufacturing limitations. The enhanced complexity of geometric design enabled by AM is heavily promoted in the media. Whilst these endeavours inspire creative solutions, the emphasis of marketing campaigns, which portray optimum design efficiency, can give a sense that the field is more developed than it actually is. Conversely, in research literature the limitations of AM are well documented, and the associated design constraints are material, machine and process-specific (Thompson et al., 2016).

Thomas-Seale et al. (2018) demonstrated that education underpins the propagation of AM knowledge in industry. Thus, a sustainable solution to this barrier would be the integration of DfAM into undergraduate engineering programmes. However, the lack of comprehensive DfAM education in most undergraduate programmes is well-acknowledged, further requiring specialist education programmes to equip engineers with the DfAM skills (Ford & Despeisse, 2016). A recent review (Ford & Minshall, 2019) further identified that AM education research is disjointed, spanning multiple disciplines and pedagogical environments. For the purposes of this study, focus shall primarily be given to the critical analysis of engineering education research undertaken in a higher education (HE) setting.

The advantages and opportunities of teaching DfAM are well documented; for example as a prototyping technique (Carfagni et al., 2020; Pieterse & Nel, 2016), to teach AM fundamentals (Go & Hart, 2016) and for subject specific applications (Horowitz & Schultz, 2014). Several studies outline frameworks and strategies for implementation (Go & Hart, 2016; Stern et al., 2019) with a light-touch qualitative analysis of the participants outcomes. Studies with a higher pedagogic focus have quantified the positive outcomes (from the students’ perspective) of integrating DfAM: supported innovation and learning motivations (Chiu et al., 2015), ideation (novelty and quality) (Hwang et al., 2020) and increased motivation, interest and ease of learning (Minetola et al., 2015). Chekurov et al. (2020) quantitatively assessed creativity resulting from a DfAM course, and the increase in creativity over 5 consecutive years. This study demonstrated a measurable increase in creativity through development of the course content, notably: an experimental AM community at the university, the increased presence of AM in the media and a physical part-library (Chekurov et al., 2020).

Fernandes and Simoes (2016) focussed specifically on learning styles and showed that the students ranked (unanimously highly) the importance of real-life 3D printing in teaching. More recently, Prabhu et al. (2020) explored the impact of teaching restrictive (e.g. geometric limitations) versus restrictive and opportunistic (e.g. enhanced geometric complexity) DfAM concepts on student creativity. Whilst the change in teaching content did not affect the students’ creativity (uniqueness and usefulness), nor their self-evaluation; teaching both restrictive and opportunistic design resulted in designs with higher “technical goodness” (Prabhu et al., 2020). To date, the outcomes of different teaching approaches for DfAM, on technical and participant perspective outcomes, have not been measured. This information is important, as it can guide academic faculty towards the most efficient and effective method of implementing DfAM into undergraduate programmes. This study shall address this gap in pedagogic research literature.

The study shall test the efficacy of and engagement with knowledge transfer of DfAM using different teaching modalities. The aim of this research is to investigate the difference in the participant’s technical aptitude and perspective after undertaking different DfAM teaching approaches; a lecture approach vs a lecture combined with a kinaesthetic and problem-based learning (PBL) laboratory approach.

Theory

Research hypothesis

With the view to integrating DfAM into the HE engineering syllabus, the foremost question is to identify the most efficient and effective method of doing so. Ideally, a full perspective of DfAM would be gained from a comprehensive and multi-modality learning environment. However, to incorporate the full breadth and depth of DfAM including material science, computation and machine design is a huge challenge (Go & Hart, 2016). This study will focus on DfAM related specifically to mechanical design and design for manufacturing, as would be covered in a traditional undergraduate mechanical engineering programme.

It is well acknowledged in the pedagogical literature, that a range of teaching strategies are required to engage with all students (Smith et al., 2005). The aim of capturing a diversity of students (in terms of motivation, attitudes, and response) requires consideration of learning styles, approaches to learning and intellectual development levels (Felder & Brent, 2005). This study will focus on learning styles, as a lens through which teaching approaches can be most easily adapted. However, it should be noted that a student’s approach to learning and their intellectual development, may also be something inherent to the student, or something influenced through the educational environment. These factors may also indirectly affect the outcomes of this study.

When considering engineering in HE, the modality of information, is very important. Engineering, whilst widely known for its mathematical and computational emphasis, remains a discipline which is highly focussed on spatial parameters for example: the breakdown of forces within a system, the three-dimensional modelling of a part, and the geometric implications of manufacturing on design. The importance of a “learning by doing” experience, such as a mechanics laboratory, is irreplaceable.

The research into the styles of learning is extensive (Coffield et al., 2004). Of these, the Dunn and Dunn model focuses on how methods of concentration and learning difficult information varies between learners (Dunn & Dunn, 1974; Jonassen & Grabowski, 1993). There are four learning modalities in the Dunn and Dunn learning style: auditory, visual, read–write and kinaesthetic (Jonassen & Grabowski, 1993). Considering these learning types, and with reference to the “learning by doing” experience for engineering students; the importance of kinaesthetic learning, is clear.

The positive impact of active or PBL is often the subject of pedagogical research. The well-known definition of Bonwell and Eison (1991) states that “active learning requires students to do meaningful learning activities and think about what they are doing”. Dym et al. (2005) highlight the importance of project-based learning “as one of the more effective ways for students to learn design by experiencing design as active participants”. Yet, neither definition specifically references a kinaesthetic approach. In the pedagogic literature specifically for design and engineering, the concept of a hands-on learning approach, is often associated with the terms active learning or project (or problem) based learning). Indeed, the positive impact of hands-on, active, PBL (and similar active learning environments) in education have been reported extensively, in STEM (Freeman et al., 2014) and engineering (Prince, 2004).

Using 3D printing to learn about other subjects, in the sense of being able to rapidly prototype artefacts, has been widely regarded as a positive teaching resource (Ford & Minshall, 2019). To date, the pedagogic literature on DfAM also reports a positive increase in student experience and outcomes (see introduction). However, teaching approaches in the university environment, are inherently constrained by resources and time, and therefore lectures have long persevered as the predominant modality for teaching. To establish whether the integration of a kinaesthetic PBL teaching approach (over lecture-based teaching in isolation) improves student learning outcomes, an exploration with reference to specific teaching approach is required. This study hypothesises that reinforcing DfAM teaching by incorporating a kinaesthetic PBL approach alongside the traditional lecture-based modality, will increase the technical aptitude of student participants. In addition, this study will measure student engagement and whether any student preferences were shown to a learning approach, and whether this was affected by the order of which the teaching session was undertaking (lecture then laboratory vs laboratory then lecture).

This study was designed to answer the following research questions (RQ):

RQ1) Does the inclusion of kinaesthetic PBL alongside lecture based DfAM teaching, increase the technical aptitude of student participants, against measures of design function, design for manufacturing, creativity and manufacturing?

RQ2) Do students show a preference towards a teaching approach and is this preference affected by the order in which the different teaching sessions were undertaken?

Study design

This study will compare the efficacy of knowledge transfer in a traditional lecture-based environment, and a lecture reinforced by a hands-on, PBL, laboratory approach. With respect to learning modalities, these two approaches represent an auditory and visually driven method (lecture) compared to a kinaesthetic focussed method (laboratory). Henceforth the teaching styles will be referred to as the lecture and laboratory format.

In a high student volume educational setting, accessible AM platforms are likely to be limited to polymer prototyping. For example, the common and inexpensive, fused deposition modelling (FDM) 3D printers. Interaction with an FDM platform, can offer students an awareness of the interdependency of materials and manufacturing parameters in a time and cost efficient manner. In turn, reinforcing the requirement to seek this information early on in the design process. Thus, the laboratory teaching approach was designed utilising FDM 3D printers.

Outcome measures

The technical ability to utilise DfAM knowledge, transferred from either a lecture or both a lecture and laboratory teaching environment, was tested through a design challenge evaluated against marking criteria. The participants’ perspectives of learning, engagement, and self-efficacy for the two teaching methods were evaluated using a student questionnaire at the end of the workshop, once the lecture, laboratory and design challenge were complete. This questionnaire was predominately composed of binary and Likert style questions. With reference to the research questions (RQ1 and RQ2), the outcome measures (OM) were looking to assess the difference in student outcomes between teaching approaches.

OM1) The technical merit (geometric design, the design for manufacturing and final the manufactured part(s)), measured through a design challenge, after either a lecture or both a lecture and laboratory teaching session.

OM2) Overall student teaching preferences through a questionnaire designed to measure learning, engagement, and self-efficacy, undertaken upon the completion of the whole study.

The quality of a questionnaire that measures participant outcomes, in clinical practise and research, can be assessed using the comprehensive checklist compiled by the Consensus-based standards for the selection of health measurement instruments (COSMIN) (Mokkink et al., 2016). The psychometric properties (measured using COSMIN) of student academic satisfaction measurement tools, are summarised in the systematic review by Rahmatpour et al. (2019). The measurements of quality include; internal consistency (interrelatedness of items), reliability, cross-cultural reliability, measurement error, content validity / hypothesis testing (extent to which the scale reflect what is being measured) and structural/construct validity (extent to which the scores reflect dimensionality of the construct) (Rahmatpour et al., 2019). With respect to this study, internal consistency (often measured using Cronbach’s alpha), requires a minimum sample size of 200; in this study it could not be determined due to the sample size of n = 24. The reliability of the questionnaire was partially ascertained through one repeated question; however it could not be measured under different timings or cross-cultural conditions due logistical constraints of the study. The content validity (aim, population and expertise of investigators) and construct validity (factor extraction) of the study were examined.

Methodology

Teaching content

Twenty-four students were recruited into the study, of which the self-identifying gender ratio of male: female was 23:1. All students were registered on a BEng or MEng Mechanical Engineering programme (at the author’s institution) and had completed at least their second year of studies. The study was undertaken after the UK summer exam period (May–June). Ethical approval for the study was granted by the author’s institution. In accordance, any identifying information has been kept confidential, and the data in this study has been fully anonymised. Due to the sole female participant, and the low proportion of female students registered on Mechanical Engineering programmes at the author’s intuition, the ethical requirement to keep identifying information confidential means that the gender breakdown, will not be analysed any further.

The 24 students were split into two groups of 12. There were no criteria for splitting the students into groups; when required to work in a pair, they self-selected and they were randomly assigned to undertaken the lecture or laboratory first. Prior to both the lecture and laboratory session, some topics that were presented via auditory and visual mediums. These topics included: agenda for the course and day, health and safety, ethics, purpose of the study, overview of additive manufacturing and basic principles of FDM.

The learning outcomes of the lecture and laboratory teaching sessions were the same, but were delivered in a different way. Seven key designs constraints of FDM, the learning outcomes, and how the knowledge was translated through the lecture and laboratory environment, is outlined in Table 1 and Fig. 1. These key learning points were summarised in a hand-out at the end of the design lecture, and at the end of the worksheets for each laboratory task. In addition to this, the relevant functionality of the AM pre-build software Cura (Ultimaker, Utrecht, Netherlands) was demonstrated including: importing models, moving, rotating and scaling, customising parameters, enabling supports, brims and rafts.

Table 1 Key designs constraints, the learning outcomes, and teaching approach used in the lecture and laboratory environment
Full size table
Fig. 1
figure 1

Example of physical and computational learning aids; a part that demonstrates the dependency of cut−out shape on printing orientation, b a thermally warped part, c engineering drawing of a part with variable feature sizes, d engineering drawing of a part with variable cut-out sizes, e CAM of table-stand used to demonstrate dependency of support on printing orientation, f pre-printed topology optimised cantilever beams and g CAD of non-consolidated bending jig

Full size image

The lecture teaching content was delivered purely by auditory and visual mediums, with the teacher encouraging verbal discussion. In a small group teaching environment, discussion was possible, however this component could not be easily be scaled up to a larger scale teaching environment. The lecture content was delivered over a 2 hour period.

The laboratory teaching content was primarily administered through PBL utilising hands-on exercises. The laboratory content was delivered over a 6 hour period utilising FDM 3D printers (Replicator 2X, Makerbot Industries, USA; Duplicator i3, WanhaoUK, UK; Ingenium, Avatar 3D, UK). To begin the laboratory session, a demonstration of the 3D printer functionality was given, followed by a series of 7 manufacturing exercises, which are outlined in Table 1 to target each of the DfAM criteria. Whilst these exercises were being undertaken the workshop facilitators engaged the students in verbal discussion to support the learning outcomes. The teaching was supported by a small amount of visual and auditory media to ensure that all the content from the lecture was also mirrored in the laboratory.

Design challenge

The 12 students in each group were self-organised into pairs to complete the design project challenge. For the purpose of assessing these groups, they are denoted as follows. Groups 1–6 undertook the DfAM lecture and design project on the first day followed by the manufacturing laboratory on day 2. Groups 7–12 undertook the manufacturing laboratory on day 1, followed by the design lecture and then the design project on day 2. Thus prior to the design challenge, groups 1–6 has undertaken the lecture, and groups 7–12 had undertaken the laboratory and lecture.

The aim of the design challenge was to “redesign the interfacing brackets (Fig. 2a and b) for fused deposition modelling (FDM) so that it meets the functional dimensions labelled on the assembly drawing”, (Fig. 2c). The brackets were to be redesigned using the CAD software Solidworks (Dassualt Systemes, Paris, France). The full design challenge can be found in Appendix A. To summarise the brief, the objectives included: the interfacing surfaces, 4 holes with a specified diameter, self-supporting features, minimal support, reducing the impact of thermal warping, withstanding a horizontal load at a cut-out (A) at a specified height and minimising the mass of the part. In line with the significant geometric freedom afforded by AM, flexibility was given to the participant in terms of how they achieved this. The students submitted their design as CAD files with the orientation and support structures (including brim or a raft) set in the CAM (Cura, Ultimaker, Utrecht, Netherlands) files. The parts (including support material) were printed using a Makerbot Replicator 2X (MakerBot Industries, New York, USA) in Acrylonitrile butadiene styrene (ABS) (3D FilaPrint Ltd, Southend On Sea, UK). The dimensions of the print volume (246 mm × 152 mm × 155 mm), and a pre-set resolution in the X–Y plane (0.1 mm) and Z direction (0.4 mm) was given to the students. The other manufacturing parameters are shown in Table 2.

Fig. 2
figure 2

Baseline CAD models of the two interfacing parts to be redesigned for FDM manufacturing; a Bracket 1, b Bracket 2 and c engineering drawing of the assembly including the functional dimensions

Full size image
Table 2 FDM manufacturing parameters for ABS
Full size table

The teacher and teaching assistants were permitted to assist the students in the functionality of the software and clarification of any points in the design brief or lecture summary. However, the teacher and teaching assistant were not permitted to aid the students by applying any of the taught knowledge or their pre-existing knowledge to the design problem.

Evaluation

The design projects of all groups were evaluated against the marking scheme outlined in Table 3. An evaluation of the technical aspects of the designs was undertaken on the digital submission files (CAD and CAM) and the 3D printed parts. The designs were evaluated against the learning objectives, specifically design for additive manufacturing, and not whether the design could functionally meet the loading requirements. Each marking criteria in Table 3 corresponds to the design criteria outlined in Table 1, and additionally the success of the manufactured parts.

Table 3 Technical marking criteria
Full size table

The marking criteria were focussed on the geometric function, design for manufacturing, creativity and manufacture. The majority of the criteria (C) was evaluated through fully/partially or not met (100/50/0%) or not applicable (N/A). The manufacturing criteria C16 and C17 were evaluated as either fully or not met (100/0%). Each student was allocated the same mark for the design submitted by their group. The criteria were designed to not be subjective. The designs were double marked. The marks for each criteria were not weighted.

The participant evaluation was designed to ascertain the perspective from each student on the learning experience, engagement and their own self-efficacy. The participant evaluation questionnaire is shown in Table 4. The questions were predominantly answered through a self-assessment on a 5-point Likert scale (1—strongly disagree, 3—neutral, 5—strongly agree). Some questions were answered by choosing a modality of teaching (design lecture or manufacturing session) and two questions were answered through choosing a learning type (visual, audial, kinaesthetic, read-write).

Table 4 Participant evaluation questionnaire
Full size table

Data analysis

All data were analysed between the groups sets (1–6 vs 7–12). The participant perspective evaluation was further analysed with respect to the total participant responses. All data is displayed in terms of the mean (mark or response) and standard deviation (where appropriate). The participant evaluation data were also analysed in terms of whether the students thought their primary learning type was kinaesthetic or one of the other three. The technical evaluation was undertaken by pairs of students, and thus could not be broken down into a mark for each student and compared against primary learning type.

SigmaPlot (Systat Software Inc., CA, USA) was used to evaluate the significance of the results. The difference between technical evaluations of the designs by Group 1–6 and 7–12, against each criterion and then the total mark, was statistically analysed using the Mann Whitney U test, applicable to continuous but not normally distributed data sets.

For the participants evaluation responses, the Likert scale was analysed using the mean and (sample) standard deviation (assuming that the constructs have a linear scale), and the binary responses (lecture vs laboratory) were analysed using one-sample Wilcoxon Signed-Rank. The difference between participant’s responses between groups 1–6 and 7–12, for the Likert scale questions was analysed using the Mann Whitney U test and for binary responses using the Fisher Exact test. These statistical tests are all applicable to non-parametric data sets. Statistical significance for all tests was defined as P ≤ 0.05.

To explore the validity of the construct, the responses for the Likert scale questions (Q1-10, Q16-21, Q26-30) were analysed using factor analysis (IBM SPSS Statistics (IBM, NY, USA)). Factors were explored through Principal Component Analysis (PCA) (De Winter & Dodou, 2016); eigenvalues above 1 were extracted, and a Varimax with Kaiser rotation was applied.

Table 5 Summary of factor analysis and hypothesised themes
Full size table

Results

Technical evaluation of the design challenge

Figures 3 and 4 show the mean technical marks for each criteria and images of the computational designs and manufactured parts submitted by each group (respectively). The technical evaluation was marked against the criteria which align to the key assessment points outlined in Table 1, with an additional category denoted “design and manufacturing” to encompass the required criteria such as the part falling within build volume, file format and final manufacturing success. The raw marks for each design under each marking criteria are shown in Appendix B, Table 6. Where marks were not in agreement, the marks have been averaged between the two markers, thus the values may vary from the standardised 100/50/0% to include 75/25%. Where detail is required to clearly justify a mark, it has been included in Appendix B, Table 7. The partially met category encompasses all marks from 75–25%. The mean and standard deviation of the marks for each marking criteria and in total for the participants in groups 1–6 (n = 12) vs groups 7–12 (n = 12) are shown in Fig. 3. Where the difference is statistically significant, the P value has been included.

Fig. 3
figure 3

The mean technical mark and standard deviation (and statistical significance where relevant) between the participants in groups 1–6 and group 7–12, against each criteria and total technical evaluation mark (as a percentage of the maximum possible mark)

Full size image
Fig. 4
figure 4

The CAD and manufactured parts of Groups 1-12

Full size image

The mean technical mark for C13 (parts have been consolidated into one part), C14 (the bracket is hollowed where possible) and C16 (manufacture was successful) were statistically higher for groups 7–12 than groups 1–6. Other outcomes of note include full technical marks for all groups for C5 (the direction of load (A) is parallel to in-plane direction of print), C10 (the part volume (including supports) falls inside the build volume) and C11 (wall thickness is above the minimum size of 2 mm). The mean technical mark for all groups was below 40% for C15 (did group seek inspiration from other sources?). There was no statistically significant difference between the group sets for the total mark (%).

Participant evaluation questionnaire

The raw data for the participant evaluation questionnaire, groups 1–6 and groups 7–12, is shown in Appendix C, Table 8. The mean response and standard deviations of Q1-10, Q16-21 and Q26-20 (the Likert Scale responses), for participants in groups 1–6 (n = 12) vs groups 7–12 (n = 12), are shown in Fig. 5a–c. Question 20 (“I found the self-learning aspects of the manufacturing challenging”) demonstrated a statistically significant difference, with group 7–12 giving a higher Likert scale response, i.e. an agreement, than group 1–6. There was no statistical difference between the reponses of the participants in group 1–6 and 7–12 for any other Likert scale question.

Fig. 5
figure 5

Group 1–6 and 7–12 responses of the mean (and standard deviation) Likert scale questions for a Learning, b Engagement and c Self-Efficacy; Design and manufacturing questions for d Learning and e Engagement and Self-Efficacy; and f the primary learning type (one participant of group 1–6 identified with two learning types). The statistical significance has been included where relevant

Full size image

The responses for the design and manufacturing questions (Q11–15, Q22–25 and Q31), for participants in groups 1–6 (n = 12) vs groups 7–12 (n = 12), are shown in Fig. 5d and e. There was no statistical significance between groups 1–6 and groups 7–12 for any of these questions. The primary learning type of the participants are shown in Fig. 5f. One student identified with both visual and audial as their primary learning type. A second analysis of the participant evaluation data of participants was undertaken. This time, the participants with a primary learning type of kinaesthetic (n = 11) were analysed against participants which identified with another primary learning type (n = 13). There was no statistical significance between the responses of these groups of participants.

When the responses was assessed across the total number of participants, some questions demonstrated a difference in mean response (where the population median was assumed to be an equal number of responses for design/manufacturing) (Appendix C, Table 9). The total responses demonstrated a significant difference towards the number of participants who chose the manufacturing option in answer to questions: Q13 = 83% (P = 0.001), Q15 = 83% (P = 0.001), Q22 = 75% (P = 0.014), Q24 = 83% (P = 0.001) and Q31 = 74% (P = 0.022).

Psychometric analysis

The participant evaluation questionnaire had 33-items, of which 21-items were measured through a 5-point Likert scale. The full sample size was 24, of which 21 questionnaires were completed in full. In this study, the Likert scale questions were aimed at one research question (RQ2—do students show a preference towards a teaching approach?) and were grouped into three themes (learning, engagement and self-efficacy).

The content validity (the degree to which the Likert scale measures the constructs) was assessed by comparing the aim, target population and concepts against the measurement items, sample population and expertise of the investigators (Rahmatpour et al., 2019). The development of the participant evaluation questions was undertaken by a team of investigators with a huge breadth of research and teaching experience encompassing mechanical and design engineering, and additive manufacturing, using both lectures and laboratory teaching approaches. Therefore, it is reasonable to assume that, based on experience, that the measures adequately reflect the research question (RQ2).

Internal reliability was investigated through one repeated Likert scale question (Q1 and Q8), with slightly different phrasing. The means of the participants evaluations were similar, Q1 = 4.5 and Q8 = 4.4. However, the raw data (Tables 8 and 9 in Appendix C), show that several participants, changed their responses, between these questions.

To explore the construct validity (the degree to which the instrument is consistent with the hypothesis, (Mokkink et al., 2016)), the dimensionality was explored using principal component dimension reduction. This techniques was used to explore common dimensions between the variables. A sample size of n = 21 (number of complete questionnaires) is very small for a factor analysis, however the minimum sample size for factor analysis has been researched extensively with contradictory outcomes (de Winter et al., 2009; Mundfrom et al., 2005). The inclusion of all variables led to a nonpositive definite matrix. Four variables were removed (Q1, Q4, Q18 and Q19), due to their similarity in phrasing and participant responses (to Q8, Q3, Q17 and Q20 respectively), to eliminate the linear dependency between variables. The analysis is summarised in Table 5; it extracted 6 factors and the rotated component matrix can be seen in Table 10 of Appendix C. It should be noted, that the small sample size made the factor analysis extremely sensitive to changes in the number of variables.

The Likert scale questions were originally designed to explore the participant evaluation through three dimensions (learning, engagement and self-efficacy). However, the factor analysis shows that the results load onto 6 distinct factors. Whilst some of the original groupings of questions cluster onto similar factors, there are additional themes running throughout the participants interpretation of the questions. The hypothesised themes, based on the factor analysis, are outlined in Table 5 Factor (F) 1 is loaded by questions under the original self-efficacy theme, specifically confidence. F2 is loaded by three questions under the original theme of learning, specifically style. F3 and F4 were loaded by questions across the original themes, it is hypothesised that they represent the themes knowledge retention and interest, respectively. The variables loaded onto F5 represent learning specificity and onto F6 represent engagement in terms of independence of learning. It should be noted that Q17 did not load significantly onto any of the 6 extracted factors. Subject to the limitations of the psychometric analysis, the factor analysis has demonstrated that the majority of Likert scale questions load onto factors that relate to the original hypothesis (RQ2).

Discussion

Technical evaluation

The mean technical mark for C13 and C14, which were classified under creativity, were statistically higher for groups 7–12. This result shows that the students who completed the manufacturing laboratory and the lecture, demonstrated an increased frequency of creative features in their design. The final creativity design criteria (C15), concerned with “seeking inspiration”, had a mean technical mark for both group sets below 40%, i.e. very few participants sought inspiration from other sources. Interestingly, “creativity” as a design criterion, was not taught explicitly through either teaching format. Although it was implicit in the challenge to “redesign” the bracket, and through creative solutions demonstrated in the teaching media, the lack of explicit teaching on seeking inspiration meant that the concept did not translate efficiently to the participants.

Groups 7–12 also demonstrated a higher mean technical mark for C16, which was for the successful manufacture of the bracket. A combined lecture and laboratory approach was more effective at transferring DfAM knowledge, leading to successfully manufacturing an AM part. Of additional note, the technical marks for C5 (optimum load direction), C10 (max part volume) and C11 (minimum wall thickness) were 100% for all groups 1–12, indicating that both teaching modalities were equally effective at transferring this knowledge. These marking criteria C5, C11 and C10 were unambiguous with a clearly defined magnitude limit or direction.

Participant evaluation

The difference in the participants perspective on learning, engagement and self-efficacy with respect to the order in which they did the lecture and laboratory, or their learning type (kinaesthetic or otherwise) predominately showed no statistical significant difference. Question 20, “I found the self-learning aspects of the manufacturing challenging”, within the category of engagement, demonstrated a statistically significant difference between the responses of group 1–6 and 7–12. The mean of group 1–6’s responses (2.4 ± 0.26) were between neutral and disagree and group 7–12’s responses (3.58 ± 0.26) were between neutral and agree. Thus, the participants who undertook both the lecture and design challenge before the laboratory, found the self-learning aspects of the manufacturing session less challenging. This is likely to be attributed by the increased accumulation of knowledge prior to the manufacturing laboratory.

When analysing the total participant evaluation data, there was a statistically significant difference between the sample mean and the null hypothesis (assumed to be an equal number of responses) for some of the design and manufacturing questions. There was a bias towards manufacturing responses for questions on aiding applicability of knowledge, feeling prepared to undertake coursework, feeling active and confident to ask questions and gaining most confidence in DfAM. None of the teaching modality questions showed a response bias towards the design lecture. Feeling active and confident to ask questions during the laboratory, both attributes of engagement, can be attributed to the PBL approach. The applicability of knowledge, feeling prepared to undertaken coursework (learning) and gaining confidence in DfAM (self-efficacy) can be considered as longer term, positive outcome which (from the participants perspective) were due to the manufacturing laboratory as opposed to the design lecture.

Psychometric analysis

In this study, the internal consistency could not be established due to sample size. However, whilst widely acknowledged as a measure of validity, it has been argued that since the Coefficient Alpha measures interrelatedness between items, it is measuring consistency between items, and not explicitly demonstrating that the construct measures what it was intended to measure (Knekta et al., 2019).

The application of a Likert scale to constructs which measure students evaluation of teaching approaches is common, however, the applicability of the approach should be noted. A Likert scale evaluates the constructs in non-linear manner, thus the data can-not be considered continuous along the scale. For example, the relative difference between the scale measureing “neutral” and “agree” compared to “agree” and “strongly agree” is highly dependent on both the question itself and the participant. Furthermore, wider generalisation of the results must be undertaken with caution, and full knowledge of how the sample population reflects the target population; limitations are imposed in terms of the particularly high male:female gender bias, the single (non-repeated) experiment, and the fact that the study was conducted at only one UK institution. Finally, layout of the questions (Hartley & Betts, 2010; Nicholls et al., 2006), scale (Courey & Lee, 2021), the wording of the questions and how the participant interprets that (Gee, 2017), have been shown to create a bias in questionnaire outcomes.

Although PCA is commonly applied as an exploratory factor analysis method, it is subject to limitations (De Winter & Dodou, 2016). The most restrictive limitation in this study, was the sample size of n = 21. Whilst some research has investigated exploratory factor analysis with small samples sizes (Mundfrom et al., 2005), the widely accepted minimum is n = 50 (de Winter et al., 2009). Any generalisations drawn from analysis of the Likert scale data, would need to consider this limiting factor. Furthermore, whilst in this study, some Likert scale questions were reverse worded to avoid bias, future work would need to consider the impact of reverse wording on the factor analysis (Zhang et al., 2016).

Study limitations

The main limitation in this study was the low number of participants (n = 24), and the population from which they were drawn (one UK institution); this limits the generalisation of the conclusions. Whilst the statistical methods applied in the study are applicable to the non-parametric datasets, a small sample size, can affect the robustness of the analysis. This was particularly evident in the psychometric evaluation of the Likert scale participant evaluation, where the internal consistency could not be established, and the factor analysis used to explore construct validity was limited by the sample size n = 21. Future research would require a much larger sample size, across multiple institutions, and time points. This would increase the significance of the research by enabling the results to be generalised over a wider population, and increase consistency, validity and reliability of the participant evaluation.

The study did not include a control set, thus the mean base-line knowledge of the students is unknown and the null hypothesis (and the population mean) for the design and manufacturing questions was assumed to be equal. Finally, it should be noted that the teaching content in the lecture and laboratory environments was administered over different time durations. This was a necessity due to the increased amount of time needed to do hands-on experiments, however it gave the laboratory and lecture group a total of 8 h to process the teaching content, as opposed to 2 h for the group undertaking only the lecture prior to the design challenge.

Summary

With reference to RQ1, there was an increase in the technical merit of ‘creativity’ and ‘successful manufacture’ for students who had completed both the lecture and laboratory prior to the design challenge, compared to the null hypothesis. These technical results show that learning was reinforced for the participants who undertook both the laboratory and lecture, leading to increased knowledge transfer. With reference to RQ2, the final participant’s evaluation demonstrated that laboratory teaching modality resulted in feelings of increased knowledge applicability, preparedness and confidence in knowledge. There was no result, that indicated that design lecture in isolation led to increased technical nor participant evaluation outcomes.

The importance of PBL (project or problem) learning is well-recognised in pedagogic literature, for example Dym et al. (2005). In more recent studies the value of real-life 3D printing in teaching has been evaluated, with several studies noting the increased outcomes in creativity (Prabhu et al., 2020), novelty and quality (Hwang et al., 2020). This study contributes to the body of research knowledge in this area by exploring, more specifically, how DfAM teaching approach affects technical and participant perspective outcomes. Whilst reinforcing previous studies, which have noted the impact to creativity, this study has further demonstrated that DfAM laboratory teaching approach increases technical outcomes in terms of manufacturability and participant evaluations in terms of applicability, preparedness, active participation, confidence to ask questions and confidence in DfAM.

This study poses two questions that need to be established by future research. Firstly, the applicability of the results over a broader student population would need to be established through larger samples sizes across multiple institutions. This would also further ascertain the validity of the Likert scale questions against the research question. The second recommendation for future research would be to assess the cost against the benefit, of integrating a wider and more expensive suite of AM techniques and materials into DfAM teaching approaches. In this study two teaching approaches for DfAM were analysed. Yet, they only focussed on one FDM technique and thermoplastic material as means of cost and time efficient, knowledge translation. As described by Kolmos et al. (2016) the methodology utilised in this research, could be incorporated into an undergraduate programme through an add-on approach or integration with existing content. However, in contrast, Go and Hart (2016) describe AM as “truly multidisciplinary” and recommend that programmes embrace the breadth and depth of this educational context.

Conclusions

This study evaluated the technical merit and the perspective of 24 students undertaking a design challenge, after either a design lecture or both a design lecture and a manufacturing laboratory. The results of the study demonstrate a significant increase in technical aptitude in the areas of creativity and manufacturing success for the participants who undertook both the lecture and laboratory prior to the technical assessment. Through evaluation of all the participants’ perspectives, a higher proportion of students reported increased applicability, preparedness and confidence resulting from the laboratory as opposed to the lecture.

In summary, this research has demonstrated the importance of laboratory PBL in DfAM teaching, leading to increased technical merit in areas of creativity and manufacturability, and the student’s perspective on their learning and self-efficacy. In the context of the economic potential that AM offers to industry; this study demonstrates that teaching DfAM in HE, using a real-life laboratory approach, will result in graduates with more confidence and a higher technical aptitude, who are better prepared to enter the rapidly developing landscape of industrial AM.