Design and Implementation of a Web-Based Automatic Grading Platform for Simulink Models
Licheng Fang
Huzhou College, Huzhou, 313000, China.
Yunhan Li *
Huzhou College, Huzhou, 313000, China.
Chenzhe Zhao
Huzhou College, Huzhou, 313000, China.
Ligeng Cui
Huzhou College, Huzhou, 313000, China.
*Author to whom correspondence should be addressed.
Abstract
Background: Simulink is widely used in control engineering education for modelling and analysing dynamic systems. However, manual assessment of Simulink assignments can be time-consuming and inconsistent, particularly when submitted models contain complex differences in blocks, ports, parameters, and signal connections. A web-based, structure-aware automatic grading platform may therefore improve grading efficiency, consistency, and the traceability of model differences.
Aims: To design and implement a Web-based automatic grading platform for Simulink model assignments in control engineering education, addressing time-consuming manual grading, difficulty in locating complex structural differences, and inconsistent grading scales.
Study Design: Software system design and controlled experimental validation.
Place and Duration of Study: Huzhou College, Huzhou, China, during the platform development and controlled validation phase.
Methodology: The platform integrates a Web interface, backend task scheduling, MATLAB/Simulink model analysis, and database management. Each submitted model is compared with an instructor-provided reference model. Structural differences in blocks, input ports, output ports, signal connections, and model-level properties are classified and converted into scores across five dimensions. Twenty-one controlled model variants (M01-M21) were each submitted twice, yielding 42 grading records.
Results: Duplicate submissions produced identical total scores and identical scores across all five dimensions for all 21 variants. Parameter modifications, port additions/deletions, connection changes, and multiple structural changes were differentiated in the corresponding dimensions. As the number of added blocks increased from 26 to 101, the total score decreased from 94.2 to 85.6. The unchanged baseline and connection-rerouting cases did not fully match theoretical expectations, indicating that the reference-model version and difference-classification rules require further calibration.
Conclusion: The platform integrates Simulink structural comparison, automatic grading, result persistence, and online querying in a repeatable and interpretable workflow, providing a practical implementation for automated assessment of model-based coursework.
Keywords: Simulink, automated assessment, model comparison, structural difference analysis, multidimensional scoring