Search

Due to the high labour costs in industrialised countries and the increasing demands on the quality of plastic moulded parts produced by the injection moulding process, companies are challenged to increase their production efficiency. Data-based approaches provide an opportunity for systematic and objective process set-up and control. Especially the process set-up can be a tedious, iterative and complex task which can consume a significant amount of time in the preparation of mass production. Artificial neural networks (ANN) can be used by evolutionary algorithms to search for suitable machine setting parameters for the desired part and process quality features. This often results in locally optimal solutions, which do not guarantee the best possible machine settings. Invertible neural networks (INN) could be a solution as with the help of these models, globally optimal solutions may be identified for definitive optimisation as the entire solution space for given quality parameters can be mapped in short time. In our approach, data for the implementation of the INN is collected in practical tests. The resulting weight and several geometric dimensions of the parts are determined as quality parameters. Based on these data, ANN are trained and utilized for the generation of synthetic data for the training of the INN. The creation of synthetic data is necessary because INN have millions of trainable parameters and therefore degrees of freedom which need to be determined using a bigger amount of training data in comparison to traditional feedforward ANN. In this paper, two different databases with a size of 10,000 samples each are being investigated. The two databases differ in the distributions of holding pressure and cooling times in order to investigate the influenceability of the prediction quality of the INN by training data. The sample settings for the injection velocity, packing pressure, mould temperature and melt temperature are uniformly distributed. Two different types of conventional and conditional invertible neural networks are implemented based on literature recommendations. Measured on validation data, the model quality (R² value) for the prediction of the part weight is 0.995 and 0.975 for the part dimensions. The machine setting parameters proposed by the INN can be analysed in regard to their impact. The holding pressure time is identified as the machine setting with the greatest influence on both part weight and part dimension. The predictions of the INN are validated in practical experiments with an R² value of 0.997 for the part weight. The weight of the parts produced with the proposed machine settings deviates on average by 0.24 % from the target. [ABSTRACT FROM AUTHOR]

The convoluted and dynamic nature of cancer growth makes it difficult to predict long-term cancer survival. Evolutionary deep learning techniques have been growing in popularity as efficient and accurate prediction techniques in recent years. Long-term cancer survival prediction has been developed to utilize evolutionary deep-learning techniques. This research combines the power of evolutionary algorithms using Particle Swarm Optimization (PSO) with some deep learning techniques to improve the performance and accuracy of the deep learning models by optimizing its parameters and minimizing the error between the predicted and actual outputs. Overall, combining evolutionary algorithms with deep learning models for cancer survival analysis prediction demonstrates the potential to significantly improve the accuracy and efficiency of cancer survival predictions, leading to better treatment decisions and improved patient outcomes. This paper uses the gbsg2 survival analysis dataset, a well-known dataset commonly used in survival analysis studies. [ABSTRACT FROM AUTHOR]

The efficiency of the systems for the automated formation of artificial neural networks is beyond any disputes. Such systems make it possible to automatically generate ready-to-work models without participation of a domain expert while creating a model. The present paper considers a modification of the evolutionary algorithm for the automated formation of artificial neural networks of direct propagation, which helps build recurrent neural networks. The peculiarity of the algorithm lies in the network structure encoding in the form of a tree allowing the network structure to be represented more compactly, that well correlates with text analysis tasks, the dimensions of which vary from hundreds to thousands of features. To test the approach, the problems of classifying textual information were solved. The averaged test results as well as some of the obtained models are presented to demonstrate the approach. It is showed that the algorithm successfully builds models with a recurrent structure. This approach can be potentially improved by adding various recurrent links. [ABSTRACT FROM AUTHOR]

The Hybrid Flow Shop Scheduling (HFS) topic has received wide attention regarding problem modeling and solution techniques. The HFS problem has been extensively researched due to its importance and complexity. This paper presents an optimization scheme for a case study on minimizing makespan and energy consumption in a hybrid flow shop scheduling problem. The HFS consists of a few productions line with several parallel machines in one or more stages. The significant difficulty is assigning work to specific devices at different locations. A case study was implemented with twenty-seven jobs in four stages comprising bending machines, lathe machines, milling machines, and wire cut machines. The (EE-HFS) optimization has been carried out using Multi Objectives Dragonfly Algorithm (MODA). The optimization result was compared with well-established algorithms, the Pareto Envelope-based Selection Algorithm II (PESA2), Multi Objectives Evolutionary Algorithm Based on Decomposition (MOEA/D), and new algorithms, Multi Objectives Grasshopper Optimization Algorithm (MOGOA) and Multi-Objective Ant Lion Optimizer (MOALO). Five metrics were employed for comparison purposes, Error Ratio (ER), Spacing (SP), Maximum Spread (MS), computational speed (CS), and Generational Distance (GD). The MODA outperformed other algorithms based on the convergence metric and was on par with different algorithms for spacing and maximum spread metrics optimization results. In conclusion, these results are significant for Hybrid flow shop scheduling regarding the energy utilization model and MODA's excellent potential to be implemented further in other combinatorial scheduling problems. This case study considerably benefits the company by revising its time and minimizing its daily energy use. [ABSTRACT FROM AUTHOR]

Energy load forecasting is a systematic method of predicting future loads in advance to maintain a balance between energy demand and supply. Sustainable energy development is critical in energy design, planning, generation, transmission, and distribution, which leads to computational complexities in conventional methods when it is non-linear. Bio-inspired optimization algorithms are part of artificial intelligence that is inspired by the biological behavior and evolutionary principles of nature. Due to advancements in artificial intelligence, an increasing number of heuristic optimization techniques have resulted in the efficient handling of complex problems. The complexities in the accurate prediction of electricity can be achieved with evolutionary algorithms and swarm intelligence optimization algorithms. It avoids unnecessary expenditure, overestimation, and underestimation of electric energy and maximizes capacity utilization. This paper reviews recent studies on population-based optimization techniques to handle complexities in energy load forecasting and smart energy management. [ABSTRACT FROM AUTHOR]

Inventory models for stock-dependent demand have been developed continuously in recent years. In this paper, we develop inventory dependent model for multi-items by considering stock-dependent demand and stock-dependent holding rate for a rack that has a specific capacity. Two models are developed. The first model is an individual order quantity for each item and the second model is a joint replenishment model. The non-linear model with constraint is solved using an evolutionary algorithm. A numerical example and sensitivity analysis are conducted to show how the models work and to get some management insights. The sensitivity analysis shows that there is no dominant model. However, the joint replenishment order outperforms the individual order when the rack capacity is tight and there is a high ordering cost [ABSTRACT FROM AUTHOR]