Academia.eduAcademia.edu

Figure 3 – uploaded by Lorenzo Salas-Morera

See full PDFdownloadDownload figure
Figure 3: Relationship between material flow and Djnf parameters for the benchmark problem O7.

Figure 3 Relationship between material flow and Djnf parameters for the benchmark problem O7.

Related Figures (18)

Figure 1: Phases of the basic CRO algorithm.
Figure 1: Phases of the basic CRO algorithm.
Figure 2: Facility layout chromosome example.
Figure 2: Facility layout chromosome example.
Fitness computation proceeds as follows:
Fitness computation proceeds as follows:
Figure 6: Example of a facility and its corresponding symmetrical layouts.
Figure 6: Example of a facility and its corresponding symmetrical layouts.
Figure 7: Flow diagram of the Interactive Coral Reef Optimization Algorithm.
Figure 7: Flow diagram of the Interactive Coral Reef Optimization Algorithm.
Figure 8: Material flow requirements for the Slaughterhouse problem.
Figure 8: Material flow requirements for the Slaughterhouse problem.
Figure 9: Material flow requirements for the CartonPacks problem. Table 3: Facility features for the CartonPacks problem
Figure 9: Material flow requirements for the CartonPacks problem. Table 3: Facility features for the CartonPacks problem
Figure 10: Material flow requirements for the ChoppedPlastic problem. Table 4: Facility features for the ChoppedPlastic problem
Figure 10: Material flow requirements for the ChoppedPlastic problem. Table 4: Facility features for the ChoppedPlastic problem
In this section, the results obtained by the proposed algorithm in the aforementioned problems are presented and discussed. ‘Tables 6, 7 and & depict the following information per algorithm’s run: the first section is re ferred to the first solution found that was given an evaluation of 5 by th DM; the first column in the section corresponds to the generation when tha: solution was found, followed by that solution’s layout, Material Handlins Cost (MHC), number of unfeasible facilities and fitness value. The seconc section is referred to the best solution found during the whole execution This solution is considered to be the one that has the lowest fitness valu while maximizing the user’s score, whether is directly obtained from the DI (integer values) or derived by the algorithm (real values). For example, it the fifth row of Table 6, column Eval, the value is 4.999, since this partic ular solution has not been directly evaluated by the DM, but given a scor according to its similarity with other solutions from the user-evaluated set The information displayed in this section is, per column, generation whe that solution was found, layout, MHC, number of unfeasible facilities, eval uation value, and fitness value. The best solution found for each problen instance set has been highlighted in boldface. An interesting fact that car be observed in the tables is that the derived user evaluations are correct the solutions with high evaluation values (more than 4.9) fulfill the DM’: conditions, although they cannot be given a 5 because they have not beer showed to the user, since they are not cluster centers.
In this section, the results obtained by the proposed algorithm in the aforementioned problems are presented and discussed. ‘Tables 6, 7 and & depict the following information per algorithm’s run: the first section is re ferred to the first solution found that was given an evaluation of 5 by th DM; the first column in the section corresponds to the generation when tha: solution was found, followed by that solution’s layout, Material Handlins Cost (MHC), number of unfeasible facilities and fitness value. The seconc section is referred to the best solution found during the whole execution This solution is considered to be the one that has the lowest fitness valu while maximizing the user’s score, whether is directly obtained from the DI (integer values) or derived by the algorithm (real values). For example, it the fifth row of Table 6, column Eval, the value is 4.999, since this partic ular solution has not been directly evaluated by the DM, but given a scor according to its similarity with other solutions from the user-evaluated set The information displayed in this section is, per column, generation whe that solution was found, layout, MHC, number of unfeasible facilities, eval uation value, and fitness value. The best solution found for each problen instance set has been highlighted in boldface. An interesting fact that car be observed in the tables is that the derived user evaluations are correct the solutions with high evaluation values (more than 4.9) fulfill the DM’: conditions, although they cannot be given a 5 because they have not beer showed to the user, since they are not cluster centers.
Table 6: Result table for Slaughterhouse instance launches eee ee ee eee pn ee eee Eee | Otten ee ee ee ee iser and the minimum of unfeasible facilities have been simultaneously found In Table 9 it is shown a comparison between different approaches for thi JA-FLP instances addressed in this work, both qualitative and quantitative ‘or each instance, the first two columns correspond to the results deliverec yy the presented ICRO algorithm (best fitness found and mean fitness of th yest solutions obtained). Next, fitness value of the best solution obtainec xy previous only-qualitative approaches are listed. Note that these solution eached the maximum user score in their evaluations. Finally, the last col umn shows the best fitness values found in the literature for these instances ocusing exclusively in material flow and feasibility (quantitative optimiza ion). The conclusion is clear: the proposed ICRO approach is capable o yptimize plant layouts while keeping the DM’s constraints. That is why Car onPacks and ChoppedPlastic instances have a better performance compare o the best known solution and Slaughterhouse does not: the DM’s prefer ces are guiding the search, and if the DM’s preferences conflict with th shape of the best solution considering only MHC those values will never b eached. That is also suitable for the other two instances: in theses cases, th M’s preferences have led to find better solutions than the ones previousl: ‘considered as the “best” ones.
Table 6: Result table for Slaughterhouse instance launches eee ee ee eee pn ee eee Eee | Otten ee ee ee ee iser and the minimum of unfeasible facilities have been simultaneously found In Table 9 it is shown a comparison between different approaches for thi JA-FLP instances addressed in this work, both qualitative and quantitative ‘or each instance, the first two columns correspond to the results deliverec yy the presented ICRO algorithm (best fitness found and mean fitness of th yest solutions obtained). Next, fitness value of the best solution obtainec xy previous only-qualitative approaches are listed. Note that these solution eached the maximum user score in their evaluations. Finally, the last col umn shows the best fitness values found in the literature for these instances ocusing exclusively in material flow and feasibility (quantitative optimiza ion). The conclusion is clear: the proposed ICRO approach is capable o yptimize plant layouts while keeping the DM’s constraints. That is why Car onPacks and ChoppedPlastic instances have a better performance compare o the best known solution and Slaughterhouse does not: the DM’s prefer ces are guiding the search, and if the DM’s preferences conflict with th shape of the best solution considering only MHC those values will never b eached. That is also suitable for the other two instances: in theses cases, th M’s preferences have led to find better solutions than the ones previousl: ‘considered as the “best” ones.
Table 9: ICRO algorithm performance comparison. per generation, respectively. The other two are vertical lines and point to the generations when a solution was given a 5 score by the DM and when the best overall solution was found. As can be observed, the learning process is usually fast and, in cases where not all the preferences are satisfied, it is possible to correct the course (Slaughterhouse - Launch 1) if a solution which ulfills all preferences is shown to the DM. In this type of scenario, random arvae creation proves its usefulness. Another remarkable fact is that there is a strong negative correlation between fitness value and user evaluation: the DM’s satisfaction is critical for the approach presented in this work.
Table 9: ICRO algorithm performance comparison. per generation, respectively. The other two are vertical lines and point to the generations when a solution was given a 5 score by the DM and when the best overall solution was found. As can be observed, the learning process is usually fast and, in cases where not all the preferences are satisfied, it is possible to correct the course (Slaughterhouse - Launch 1) if a solution which ulfills all preferences is shown to the DM. In this type of scenario, random arvae creation proves its usefulness. Another remarkable fact is that there is a strong negative correlation between fitness value and user evaluation: the DM’s satisfaction is critical for the approach presented in this work.
Figure 13: ICRO Evolution per launch in ChoppedPlastic instance.
Figure 13: ICRO Evolution per launch in ChoppedPlastic instance.
are designed, and improve their subsequent performance in the real world. Figure 14: Best layout found by the ICRO algorithm for Slaughterhouse instance.
are designed, and improve their subsequent performance in the real world. Figure 14: Best layout found by the ICRO algorithm for Slaughterhouse instance.
Figure 15: Best layout found by the ICRO algorithm for CartonPacks instance.
Figure 15: Best layout found by the ICRO algorithm for CartonPacks instance.
Figure 16: Best layout found by the ICRO algorithm for ChoppedPlastic instance.
Figure 16: Best layout found by the ICRO algorithm for ChoppedPlastic instance.
Related topics:
Computer ScienceBio and Nature Inspired AlgorithmsCoral Reefs Optimization
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved