International Journal of Simulation and Process Modelling (8 papers in press)

Regular Issues

• Open Access

  Deep learning-driven simulation modelling for mine geological risk assessment integrating multi-source data ( Free Full-text Access )

  by Fuming Zhao, Chao Xie 
  Abstract: Mining geological risk assessment is crucial for ensuring production safety, yet traditional methods relying on single data sources and expert experience suffer from low accuracy and delayed early warning, often failing to capture dynamic risk characteristics. This paper proposes a simulation-driven intelligent assessment framework that integrates multi-source monitoring data including geological, subsidence, hydrological, and microseismic information - with advanced deep learning techniques. The framework is designed to dynamically simulate risk evolution processes and construct an end-to-end predictive model. Validation on public datasets demonstrates an accuracy of 91.3%, significantly surpassing the 78.5% achieved by traditional methods (p < 0.01), with high stability and generalisation ability. This process-modelling paradigm effectively overcomes the bottlenecks of information incompleteness and response delay, providing reliable technical support for geological hazard prevention and a foundational tool for intelligent mine construction, thereby supporting dynamic safety management.
  Keywords: multi-source data fusion; deep learning; mine geological hazards; intelligent assessment.
  DOI: 10.1504/IJSPM.2026.10076608
   
  

• Open Access

  Automated simulation testing for complex software environments using multi-agent reinforcement learning ( Free Full-text Access )

  by Qiuming Zhang, Jing Luo 
  Abstract: The growing complexity of software systems poses major challenges for automated testing in continuous integration and continuous delivery pipelines. This paper proposes multi-agent-based software automated testing, a multi-agent reinforcement learning framework that models testing tasks as a decentralised partially observable Markov decision process. Using the Q-network mixing algorithm, the system enables coordinated decisions across testing agents. Evaluation on a TravisTorrent-based simulation environment shows multi-agent-based software automated testing achieves a 95.2% defect detection rate - showing a 4.7% improvement over the best multi-agent baseline - while reducing test execution time to 70% of conventional rule-based scheduling. Compared with multi-agent deep deterministic policy gradient, it demonstrates a large effect size (Cohen's d = 0.89) in defect detection. These results demonstrate the framework's effectiveness in improving testing quality and efficiency, offering a viable solution for intelligent test automation in complex software environments.
  Keywords: multi-agent reinforcement learning; MARL; automated testing; continuous integration and continuous delivery; CI/CD; Jenkins.
  DOI: 10.1504/IJSPM.2025.10075279
   
  

• Open Access

  Discrete-event simulation modelling of inventory turnover under supply chain financial collaboration ( Free Full-text Access )

  by Zhifeng Qu 
  Abstract: This study addresses the critical gap between operational and financial objectives in supply chain inventory management by proposing a novel framework that integrates discrete-event simulation with deep reinforcement learning. We formulate a dual-objective reward function incorporating both traditional costs (holding, shortage, ordering) and the financial metric of cash conversion cycle. Trained and tested on the real-world M5 forecasting accuracy dataset, our model, cognitive load dynamic assessment model-proximal policy optimisation, demonstrates superior performance. Results show it achieves a total cost of 285.4 ± 8.7 (in thousands), significantly lower than state-of-the-art baselines (p < 0.01), while maintaining a 98.2% service level and reducing cash conversion cycle to 35.2 days. This result highlights the framework's effectiveness in achieving operational-financial synergy, offering a data-driven decision-support tool for enhancing both efficiency and financial health in dynamic supply chain environments.
  Keywords: supply chain finance; deep reinforcement learning; DRL; inventory optimisation; cash conversion cycle; M5 forecasting accuracy dataset.
  DOI: 10.1504/IJSPM.2025.10075344
   
  

• Open Access

  Enhancing retail decision-making accuracy through deep learning-based consumer sentiment simulation modelling ( Free Full-text Access )

  by Haoyang Qin, Enzhi Liu, Jiaxin Wang 
  Abstract: Confronted with the challenge that traditional retail decision systems struggle to quantify the impact of consumer sentiment, this paper proposes an agent-based simulation framework powered by a deep learning model integrating bidirectional encoder representations from transformers with a bidirectional long short-term memory network. This approach constructs an end-to-end consumer sentiment simulation system through the fusion of multimodal data, including textual reviews and behavioural sequences. Experiments on the publicly available Amazon review dataset demonstrate that this model achieves a sentiment recognition accuracy of 92.7%, representing a 15.3% improvement over traditional long short-term memory models. By systematically integrating fine-grained sentiment dimensions into the decision-making process, the system enabled a product recommendation conversion rate increase of 22.1% and an inventory turnover rate optimisation of 18.6%. The results robustly validate that the proposed sentiment simulation framework significantly enhances the precision and intelligence of retail decision-making.
  Keywords: consumer sentiment simulation; retail decision optimisation; deep learning; multimodal data fusion; BERT-BILSTM model.
  DOI: 10.1504/IJSPM.2026.10075690
   
  

• Open Access

  Multimodal transformer-driven consistent environment design generation simulation modelling ( Free Full-text Access )

  by Zhuo Fan, Jinqi Wang 
  Abstract: Automated generation of physically plausible 3D environments is a key challenge in digital twins, the metaverse, and robot simulation. Current methods focus mainly on visual fidelity, often overlooking functional and physical rationality, limiting direct applicability to simulation tasks. To address this, we propose a multimodal transformer-based framework for environment design generation and consistency simulation. Utilising a cross-modal attention mechanism, our model integrates textual descriptions with prior knowledge from real 3D scenes. It incorporates fine-grained physical constraint losses - including collision avoidance, support relations, and spatial accessibility optimisation - during training to explicitly model physical consistency. Experiments on the Matterport3D dataset show our method outperforms existing baselines in visual quality and layout rationality. Notably, it shows significant gains in physical consistency: collision volume is greatly reduced, and navigation success reaches 89%, affirming high simulability and practicality of the generated environments.
  Keywords: multimodal transformer; environment design generation; physical consistency; simulation modelling; microphysically constrainable.
  DOI: 10.1504/IJSPM.2026.10075345
   
  
• Simulation-driven deep learning framework for early poultry disease detection using faecal image classification with optimised CNN architectures  
  by Nonita Sharma, Monika Mangla, Manik Rakhra, Baljinder Kaur, Raj Kumar Mohanta, Bijay Kumar Paikaray 
  Abstract: The present investigation aims to devise an optimized Convolutional Neural Network (CNN) framework to identify prominent poultry diseases based on faecal images. The proposed research model uses a tri-convolutional layer architecture to enhance the accuracy of poultry disease classification and improve feature extraction. Three major diseases, namely coccidiosis, salmonella, and Newcastle, have been considered. The prime objective of current research is to achieve early detection by employing advanced deep-learning techniques. The proposed model uses fecal images to identify pathological conditions using the TriConvLayer architecture accurately. In this work, custom CNN models, viz. SoloConvLayer, TriConvLayer, and FiveConvLayer models are used to achieve an accuracy of 97%, 98%, and 98%, respectively. The achieved result advocates the efficacy of the proposed approach. It thus has the potential to revolutionize early disease detection in poultry farming, a major step towards improving animal health and farm productivity.
  Keywords: deep learning; poultry disease detection; convolutional neural networks; CNN; faecal image analysis; poultry health monitoring; disease classification; simulation framework.
  DOI: 10.1504/IJSPM.2025.10073525
   
  
• A simulation-based approach to multi-crack creep damage analysis in oil shale using Voronoi tessellations  
  by Shiwei Hou, Xunqing Lv, Chao Guo, Junyan Han, Zhongheng Liao, Ping Fan 
  Abstract: This study combined experimental and numerical simulations to investigate the long-term stability of open-pit mine slopes subject to coal-bearing oil shale creep failure. To establish a rock multi-crack platform for numerical analysis, we used ABAQUS driven by Python to develop the Voronoi tessellation making program (Voronoi_ABAQUS_Py) and cohesive element insertion programs (Coh_3D_ABAQUS). An improved Burgers creep model with nonlinear viscosity was proposed and implemented via a UMAT subroutine for the multi-crack FE analysis. Indoor multi-stage loading creep tests assessed damage under different stress ratios. A comparison of the experimental data with the model data proves the effectiveness of the proposed model on the multi-crack numerical analysis platform (Python-ABAQUS). The findings reveal a dual-phase damage mechanism in the long-term creep damage process of oil shale slopes is a competitive process between microcracks and the matrix. The research provides a scientific method for predicting the long-term stability of fractured rock masses.
  Keywords: Voronoi-based modelling; numerical simulation; multi-crack creep analysis; oil shale creep damage; improved burgers model.
  DOI: 10.1504/IJSPM.2026.10076616
   
  
• Actor-based simulation framework for scalable verification of mutual exclusion algorithms  
  by Libero Nigro 
  Abstract: This work develops a formal method for modelling and verifying mutual exclusion algorithms based on timed automata and Uppaal. Being based on model checking (MC), the approach is not scalable in the number N of processes due to the well-known problem of state explosions, which arises, e.g., for N ≥ 4 processes. Alternatively, the statistical model checker (SMC) of Uppaal can be used, which emulates concurrency and action non-determinism by stochastic behaviour. Although SMC permits studying a model for greater values of N, scalability problems can still be present due to data accumulation during the simulations. In this paper, a more efficient technique is proposed that is based on a lightweight, efficient, and control-based actor system. A Uppaal model is translated into Java with processes that are mapped onto actors and atomic actions that are realised by messages. The paper demonstrates the effectiveness of the actor-based approach through different case studies.
  Keywords: mutual exclusion algorithms; atomic/non-atomic registers; anonymous memory; model checking; timed automata; Uppaal; state explosions; discrete-event simulation; actors; Java.
  DOI: 10.1504/IJSPM.2026.10076157