E4a | UML CD/SC vs. SC/CD order Vogel-Heuser, Seidel [78] | N | ~ | (p&p) structure/ behavior DI/O 13 (DI 9; DO 4), WMC = 11 small group (0.42) | L (3.2) E (0.42) | BSc 2.Sem 41CD/21SC | - | no significant differences in terms of modeling order |
| Forcing subjects to follow a specific modeling order is not helpful to improve structural models; subjects had problems to create suitable classes from similar objects |
E4b | BSc 2.Sem 70CD/32SC | + | “behavior first” group: x̅ 23.4 p (SD =10.326; SE = 0.982), Structure first group x̅ 18.4 p (SD = 8.220; SE = 1.825) out of 46 p max. s~: quality structural model bad: “structure first” x̅ 12.26 p “behavior first” group x̅ 11.14 p, structural quality of model is independent from modeling order | Post experiment interview and questionnaire to reveal challenges and reasons for mistakes ~ | ||||||
E5 | UML Embedded CD/SC vs. IEC FBD Vogel-Heuser et al. [63] | N | ß | (p&p) with reuse of buffer, structure/ behavior DI/O 38 (DI 24; DO 14), WMC = 44 (UML) FBD 64 variables (2.2 + 0.92 quest.) | HLE and HB Fade out (Rep) (6.42) | 18 Apprentices 1./2. year | ++Strong correlation with abilities | Behavior (p = 0.22) and structure (p = 0.15) not significant; UML: thinking breaks Relatives to abilities (Table 1, 51): grade mathematics: (r = 0.234, p = 0.055); grade automation: (r = 0.327, p < 0.005) grade mechatronics: (r = 0.327, p < 0.005); cognitive demand: (r = −0.255, p < 0.05); previous knowledge: (r = 0.485, p < 0.001) | Frustration levels of UML group signif. higher than FBS group (p = 0.02), clearness of FBD signif. higher than of UML (p = 0.017) ease of use for behavior programming UML signif. lower than FBD (p = 0.012), Subjective quality estimation of UML group matches the factual quality signif. better than the FBD group (p = 0.025) | Difficulties with abstraction and building correct classes, e.g. cylinder, relationship CD and SC |
CG | ||||||||||
E6 | SysML-AT Parameter Diagram (PD) vs. CFC vs. IEC 61131-3 ST; Frank, Schütz/Obermeier, Schütz [25] [26] | N | ~ | Physical laws, structure, Maintenance task, analysis, understanding, interpret 4 - 5 sub-blocks, 7 - 8 variables, WMC = 5 - 6 (0.167 per notation + 0.083 quest.) | L (0.167) E (0.084) per notation | 6 BSc. (mech. eng.) All notations different sequence | Q | Correct solutions PD 68.2% CFC 63.4% ST 64.3% | Answers correlated with objective results, learning effects across notations | Adequate method for requirement analysis and architecture design was missing |
E7 | SysML-AT vs. IEC 61131-3Notation + method Frank et al. [79] [80] | N, M | Prot. | NAS Coking plant structure, hyb, communication, RT: Material sorting, belt synchronization, level control; DO 3, AI 8; AO 4 WMC = 18, (a) 2,3; b) 2,6; c) 2,6); all mean times | L (1.5) E (1.5) | 15 BSc (mech.eng) | ++ | CFC x̅ 93.5 (182) points NM x̅ 123.1 (182) points (p < 0.001) | From NM to NMCP: mental demand and workload: ¯, fear of failure: , fatalistic externality: , NMC worst values: suitability for task and individualization (see Table 3) | Compromise between support and complexity. Difficulties with module abstraction |
SysML-AT a) plus characteristics Frank et al. [81] | N, M, C | L (2) E (2) | 5 | Q+ | NM x̅ 97.20 (182) NMC x̅ 116.6 (182) | |||||
SysML-AT b) plus pattern Eckert et al. [82] | N, M, C, P | L (2.5) E (2.5) | 5 | Q- | NMC x̅ 116.6 (182) NMCP x̅ 119.75 (182) |