In this paper, we present maintenance cost optimization techniques under reliability constraints based on evolutionary algorithms for complex systems in the context of the Reliability-Centered-Maintenance (RCM). Our main goal is to find the best maintenance policy for this system by minimizing the maintenance cost function of the system, under the constraint of the required reliability for a given period. This policy identifies the optimum times in which the components must perform preventive maintenance (PM). The maintenance cost can be considered as a PM, an unscheduled maintenance (UM) and a replacement maintenance (RM) cost. The PM action is, therefore, considered to have an imperfect effect on the component in this work. The imperfect PM is executed whenever the component reliability reaches a certain threshold. The proposed method allows us to find the reliability threshold R_{\mathit{th}}^jfor each system component j related to the number n of PM actions performing the RM action on the components in order to minimize the expected total maintenance cost of the system over the mission time{(\mathbf{}T}_{\mathbf{mis}}). Therefore, the optimal PM intervals durations obtained for each component T_i^j (i = 1, 2,?, n) correspond to the optimal reliability threshold. In this study, we use the evolutionary algorithm to find the optimal reliability threshold R_{\mathit{th}}^japplied on each component j. A comparative study is performed to evaluate the performance of the Lévy-flight firefly (LFA) algorithm and particle swarm optimization (PSO) algorithm in finding the global optimum.

The purpose of data mining is to process raw data and extract rules. The Vehicle Routing Problem is a logistical problem, which handles the delivery and the collection of products. In the classical problem, the position of the depot and customers are known in advance. In case of the base problem, the demand of the customer is also known in advance. But, we may need some future data, for example, the demand of the customer, so we need to forecast these data from the previous data. After the determination of the future demands of the customers, we determined whether it is worth serving customers or not with the help of the classification methods. After the time-series forecasting and classification, we also determined the route of the vehicles with the help of the genetic algorithm.

Topology optimization is a commonly used method to determine the optimal design of damping material layout in train, automotive, and aerospace products. In this study, a coupled system of a flexible face with closed acoustic cavity is established. A unit load is applied to the flexible face, and the response sensitivity and modal participation factor of the filed point are analyzed. Three response sensitivity peaks are found at the low-frequency band (20-180Hz), and the corresponding main modal participation factors are the first, fourth, and eighth orders. The three peaks are reduced by damping treatment. A method for objectively calculating the modal weight of each order using three response sensitivity peaks is proposed. The comprehensive modal loss factor of the mode corresponding to the three peaks is targeted, the position of the damping material is the design variable, and the amount of damping material is the constraint. The evolutionary structural optimisation (ESO) method is used to design the layout of damping materials. Aiming at the checkerboard problem in the optimization process, the calculation sensitivity is filtered to stabilize the optimization. Results show that the response sensitivity of the third peak in 173Hz is reduced by 33.2 and 26.5 dB(A) under free damping and constrained damping treatments, respectively. The two other response sensitivity peaks are also remarkably decreased.

In order to achieve personalized learning in an online learning environment, this paper proposes a personalized knowledge map recommendation algorithm based on interactive behavior preferences through collaborative analysis of learners' online interactive behavior data. First, it defines the learners' interaction degree of knowledge points based on the interactive behavior in the online learning process, the learner's mastery degree of knowledge points based on the online test scores, and the learning effect in combination with the interaction degree of knowledge points. Secondly, this paper proposes a correction factor based on the stability of difference in the interaction degree of knowledge points, combined with the average interaction degree of knowledge points to improve the similarity calculation model. According to the MAE and MSE evaluation indexes of similarity prediction, the prediction effect of the similarity calculation model is evaluated. Finally, combined with the learning effect and similarity calculation model, the knowledge map is recommended for the target learners. The validity of the recommendation is proven by the F1 value and MAE evaluation index.

Technology changes life. Recently, state-of-the-art of deep learning has inspired the development of deep-learning based approaches for outfit compatibility. Constructive and useful advice on clothing matching is provided to people by outfit compatibility technology. Nevertheless, there are still some shortcomings to be overcome. To be specific, outfit compatibility is a subjective method to evaluate the effect of cloth matching because of economic factors such as price, but price factor has not been taken into consideration in previous works. In addition, the relationships among vision, text, and price features are complex. Therefore, a multi-modal input graph neural network (MI-GNN) model is proposed to solve these shortcomings. This model can not only better capture the complex relations between various items in an outfit, but also model outfit compatibility from multiple modalities. In fill-in-the-blank task and outfit compatibility prediction, the performance of compatibility modeling improved the performance by 0.96% and 0.56%, respectively.

With the rapid development of mobile Internet, micro-blogs have been already integrated into people's work and family life. Traditional topic-based semantic community detection methods have not considered the sentiment tendency of users towards specific topics and therefore increase the probability of community fragmentation. To solve this problem, we propose an event-based community detection framework in the micro-blog network. Firstly, the major events in micro-blog within a sliding window are extracted. Then, the user's sentiment polarity of each event is used as the initial community label of each individual. Finally, the community detection procedure can be implemented by iteratively updating the community label. The simulation results indicate that the proposed method can capture the consistency of views of members within the same community, which brings higher accuracy and sentiment cohesion.

Non-Intrusive Load Monitoring technology is the most common method for power system decomposition recently. In order to solve the problem of low precision of current non-intrusive load decomposition, an improved sliding window bilateral accumulation and event detection algorithm is proposed in this paper. The random forest and deep neural network and support vector machine algorithm are used to decompose and identify the non-intrusive load equipment group. The results of the case analysis show that this algorithm is significantly better than the traditional non-intrusive load decomposition algorithm for non-intrusive load equipment group decomposition.

An improved bird swarm algorithm (IBSA)-based task scheduling strategy was proposed to solve the problem of improved task scheduling and high equipment energy consumption in the cloud computing environment. Bernouilli shift chaotic mapping was used to initialize the population and improve its diversity. Levy flight feature was used to update the individual position to avoid the algorithm falling into local optimization. The bee colony algorithm was used to select the optimal individual for the next iteration. Simulation results showed that the time and energy consumption of the proposed algorithm were optimized under four different tasks. Compared with the ant colony algorithm, particle swarm algorithm, and IBSA, the algorithm presented in this paper has obvious advantages in scheduling effect and can effectively save time and reduce energy consumption.

The traditional monitoring technology of energy storage power stations has problems of poor control effect and low dispatching accuracy. Based on this, this paper studies the monitoring technology of energy storage power stations based on the discharge control scheduling algorithm. From the aspects of discharge control, daily maintenance, discharge process monitoring and power utilization rate of the energy storage station, the precise control of the discharge process is realized. The discharge control scheduling algorithm of the energy storage station is used for comprehensive analysis and evaluation. The algorithm can realize the dynamic record of multiple data entries in the discharge process of the energy storage station and realize diversified analysis and intelligent matching. Through the discharge control scheduling algorithm of the energy storage power station, the whole process monitoring of the energy storage power station is realized, so as to improve the working efficiency and reduce the cost of the discharge control link. The experimental results show that the monitoring technology based on the discharge control scheduling algorithm has the advantages of high control accuracy and fast response, which can effectively improve the monitoring efficiency of the energy storage power station.

The safe operation of the power grid has a direct impact on the stability of the power supply with the power system. Due to the large number of equipment in the power grid, it is difficult for daily patrols. The efficiency of patrols of UAVs in power grids and the optimal configuration of UAVs are the purpose of the optimal model for patrols of UAVs in power grids under time constraints. Under time constraints, the optimization goal is to minimize the flight time of all UAVs or the longest flight time of a single UAV with the condition of completing all patrol tasks. It constructs the optimal model for patrol paths of UAVs in power grids through the space-time road network. The genetic algorithm is used to solve the optimal problem of patrol paths of UAVs. The results of scene simulation show that the patrol paths of UAVs in power grids planned by this model are more in line with actual needs and have obvious efficiency advantages.

Both random testing and combinatorial testing are input-domain testing techniques. Adaptive random testing, which is an improved version of random testing, selects a test case with more differences from all the existing test cases in each step. Greedy combinatorial testing generates test cases using greedy algorithms to cover more uncovered tuple-combinations of parametric values in each step. To compare fault detection efficiencies of the adaptive random testing technique and greedy combinatorial testing technique, we design an experiment on Boolean specifications that were extracted from the TCAS system. By analyzing fault detection ratios, f-measure values, and APFD values of the two testing techniques, experimental results show that: (1) if the number of test cases is relatively small, fault detection efficiencies of the two techniques are very close though adaptive random testing has a little advantage; (2) for an increase in the number of test cases, the fault detection efficiency of greedy combinatorial testing becomes gradually better.

Prediction of software defects is an effective way to improve system quality, and it is a key factor affecting the efficiency of defect detection and repair in software components. The purpose of this study is to improve the effectiveness of component defect prediction in the following two ways: for the imbalance of training data in defect prediction and the insufficient support of single regression in predicting the number of defects in components First, this study proposed to adopt SMOTE to construct a balanced sample dataset and oversample the defective components in the unbalanced sample dataset to take into account the proportion of different types of samples and improve the accuracy of prediction; second, this study proposes a method of multi-step prediction for the number of defects that supports regression after classification, and the method applies support vector machines to classify components and filter out non-defective components in the classification results, applies regression to establish a component defect number prediction model to effectively implement the multi-step prediction of component defect number, and further improves the accuracy of prediction. The evaluation of the experiment was completed on open-source datasets. The results show that the accuracy of multi-step prediction is better than the prediction by regression alone, and multi-step prediction has higher overall efficiency and applicability.

In order to improve the efficiency of task scheduling under Cloud computing, this paper proposes an optimized Improved Particle Swarm Optimization and Membrane Computing (IPSOMC). First, it describes the cloud computing task scheduling model with time and cost as the main research object. Second, it uses Kent mapping to initialize the population and introduce domain particles to improve the global optimization ability of the particle swarm. It also uses the weight factor and nonlinear extreme value disturbance to improve the local optimization ability of the particle swarm. Finally, the optimal solution of the particle swarm algorithm is selected with the help of the evolution rules of membrane computing. The simulation experimental results show that the IPSOMC algorithm and the comparison algorithm have good effects in terms of completion time and consumption cost under different task scales and improve the efficiency of task scheduling.

In this paper, the joint connection of epoxy resin concrete truss structure is analyzed by computer simulation and ABAQUS finite element simulation. Figureon software is a set of powerful finite element software for engineering simulation that can solve problems ranging from relatively simple linear analysis to many complex nonlinear problems. It can analyze complex solid mechanics and structural mechanics systems, and it can especially handle very large and complex problems and simulate highly nonlinear problems. Fabricated joints are required for truss structure joints because the construction technology and the stress state of the truss structure are relatively complex. Through modeling multiple groups of joint models, the technical difficulties in establishing finite element models, the determination of grid division and boundary conditions, and the finite element analysis and result comparison of truss joints under various parameters are solved.

With the increasing popularity of Internet, e-commerce and environmental awareness, more and more manufacturers have chosen to produce and sell green food through a dual-channel supply chain (DCSC) to improve market and environmental performance, i.e., the demand, price, profit as well as greenness level. However, the implementation of online channels has triggered substantial changes in consumer behaviors, which arises fairness concern of the retailers and even manufacturers regarding price and profit distribution of the green food DCSC. Therefore, how to explore the influence of fairness concern on the interaction between the performance of green food DCSC and consumer behavior is a vital issue to be solved. To fill this research gap, this study constructs three models of fairness concern (all DCSC members are neutral; the manufacturer has fairness concern; the retailer has fairness concern) to analyze the effects of fairness concern coefficient on the performance of green food DCSC. The numerical studies show that, in general, the consumers' preference for green food will promote the greenness level of the product and thus heightens the total profit of DCSC. Changes in the market share of the offline will affect performance through fairness concern and transfer price. In particular, the results further illustrate that when the green food manufacturer raises fairness concern, the profit of DCSC members as well as the profit of the whole DCSC will decrease. Furthermore, when the green food retailer increases its fairness concern, the profit of the manufacturer will decrease, while the profit of the retailer and the profit of the whole DCSC will increase. However, it is worth noting that the influence of the fairness concern on the performance including pricing and profit of green food DCSC will be reduced when the behavior of fairness concern rises to a certain extent.