Abstract:

Abstract:

Abstract: To improve the adaptability of echo state network (ESN),an optimization method based on mutual information (MI) and Just-In-Time (JIT) learning was proposed in this paper to optimize the input scaling and the output layer of ESN.The method was named as MI-JIT optimization method and the obtained new network was MI-JIT-ESN.The optimization method mainly consists of two parts.Firstly,the scaling parameters of multiple inputs were adjusted on the basis of MI between the network inputs and outputs.Secondly,based on JIT learning,a partial model of output layer was established.The new partial model could make the regression results more accurate.Further,a multi-input multi-output MI-JIT-ESN model was developed for the fed-batch penicillin fermentation process.The experimental results showed that the obtained MI-JIT-ESN model performed well,and that it had better adaptability than ESN model without optimization and other neural network models.

Abstract: Principal Component Analysis (PCA) is a well known model for dimensionality reduction in data mining,it transforms the original variables into a few comprehensive indices.In this paper,we study the principle of PCA,the distributed architecture of Spark and PCA algorithm of distributed matrix from spark’s ML-lib,then improved the design and present a new algorithm named SNPCA (Spark’s Normalized Principal Component Analysis),this SNPCA algorithm computes principal components together with data normalization process.We carried out benchmarking on multicore CPUs and the results demonstrate the effectiveness of SNPCA.

Abstract: In order to meet the practical requirements of automatic application and renewal of driver’s license,a high speed system for automatic recognition of driver’s licenser was designed and implemented.The hardware was designed to capture the image of the driver’s license that contained the smallest identifiable features.Because of the complex background such as the shadow line and so on in the driver’s license images,the existing recognition algorithms had the low recognition accuracy,universality and robustness problems.This paper first solved the segmentation difficulties for uneven illumination,noise,tilt and shadow line character by combined adaptive binarization and morphological processing.Then,the Blob analysis was used to extract the important local features of the driver’s license,and the recognition accuracy was further improved by using the prior information and the correlation matching algorithm.The experimental results showed that not only the false recognition rate was 0,but also the practical products was developed,and the better social effects were achieved.

Abstract: In order to prevent the serious safety problem caused by the dry pot burning and stove explosion and firing,an anti-overheating system was designed.The system of infrared temperature sensor MLX90614 on the bottom of the pot was used to realize the non-contact real-time temperature monitoring.The real-time temperature data was collected and processed by the STM32 microcontroller and SMBus.When the temperature of the bottom of the boiler was beyond the normal heating range,the temperature monitoring module could send a voice alarm.When the threshold value of the dry burning temperature was reached,the gas circuit could be cut off by the control circuit serially connected in the thermocouple temperature detection circuit.Experimental results showed that the proposed system could cut off the gas path once the preset temperature reached and prevent the dry pot burning effectively.

Abstract: Based on the characteristic of face orientation,position and the light background in face recognition,a new method of face orientation recognition based on development network is proposed.The characteristic of human’s eye was very prominent,so the position of eyes was chosen as the face orientation feature vector.And the deveiopment network model was used to recognize human’s face orientation in the different light background images.The result showed that this method could effectively solve the difficult problem of face orientation recognition under varying illumination conditions by comparing with the test results of other methods,which was fast,stable and effective.The recognition rate was as high as 100％.

Abstract: BP neural network for effectively fusioning the information obtained by arc sensor and ultrasonic sensor and information of welding parameters such as welding current,welding speed,welding groove and so on was used to obtain the prediction model of weld penetration depth.Simulation results showed that:the prediction model of weld penetration depth could measure the weld penetration quickly,accurately and in real time.For the precise control of weld penetration,parameters self-tuning fuzzy PID controller was desing,which combined with the advantages of traditional PID controller and fuzzy controller.Smulation results showed that compared with traditional PID controller,parameters self-tuning fuzzy PID controller had a significant advantage in the performance of the system.

Abstract: The distribution of charging and battery swap station has always been one of the key problems for the development of electric vehicle.A site location of charging and battery swap station could be represented by a network with traffic flow,the distance from the power source,parcel load,and city block position respectively.Spectral clustering methodology was used to reveal the internal connectivity structure of such a network.First of all,it adopted the min-max standardized method in dimensionless to establish a sample space matrix S.Then,the normalized Laplacian was achived according to the similarity between matrix W and matrix D.The former two and three feature vector of Laplace matrix were mapped to the 2d and 3d space to observe parcel partition.Finally methodology goes beyond the standard k-means algorithm by instead representing the complete network substructure as a dendrogram and verifies its correctness by analyzing the voltage sags.We include the results of our methodology for a real distribution network in Jiangxi province.Example shows that our methodology has certain rationality and it could be helpful for distribution network planning.

Abstract: During heating process of sintering furnace,the model parameters were easy to change,and traditional PID control was difficult to achieve the desired control effect.This paper used particle swarm optimization algorithm to identify the mathematical model of sintering furnace,for sintering furnace with high inertia,time-variation and strong time delay etc,a method of supervision and control based on RBF neural network,which combined PID control with neural network control.When temperature or parameters changed greatly,PID control played a major role.neural network played a regulatory role and compensated the shortage of PID control.The simulation results of MATLAB software showed that this method could improve the control precision of sintering furnace,which had a certain practicality.

Abstract: With the increasing of highway extension projects and widely use of sheet piles in railway construction,the mechanical behavior of extension embankment was analyzed through simulating different kinds of pile and load of different positions.Then the optimal pile kind and the most unfavorable load position were proposed.Through continuous observing of settlement in sheet pile section and CFG pile section,the optimal adaptability of sheet pile was showed in extension projects.The analysis results showed that the effect on settlement of PTC pile,CFG pile and cement mixing pile was gradually decreased.The PTC pile and CFG pile should be firstly selected from the options of controlling settlement.The most unfavorable load position was in new embankment and its quality was the key control point in construction.The effect on decreasing differential settlement was appeared in process of semi-rigid base construction,and it would be even obvious in pavement construction.The sheet pile was an effective supplement to traditional soft soil treatment methods.It had better adaptability and foreground in highway extension projects.

Abstract: In order to study the impact of various parameters on attenuation law of joint load transfer efficiency of steel fiber reinforced concrete (SFRC) pavement,20 pavement specimens with different dowel bar diameter,length,thickness of concrete slab and content of steel fibers were tested in fatigue experiments.The paper briefly introduced the test arrangement including the raw materials of specimens,the parameters and production of specimens,as well as the loading system.It was shown that the steel fiber added in concrete pavement could effectively improve the load transfer efficiency and reduce the attenuation rate of load transfer coefficient.The attenuation curves of joint load transfer coefficient of SFRC specimens under different factors had similar characteristics,which could be divided into two stages:rapid attenuation and steady decline.However,the curve showd three-stage decay when the content of steel fiber was less than or equal to 0.6％.Although the increase of dowel bar diameter,length and content steel fiber could improve the joint loading performance,the increase range was extremely limited,and the impact of steel fiber content was the largest.The decline rate of load transfer coefficient,which had no relations with dowel bar diameter,length and concrete slab thickness,decreased with the increase of steel fiber content.After the load was applied 100 million times in the experiment conditions,the load transfer coefficient of SFRC specimens was remained above 90％,which showed a good load transfer capacity.The test results could provide a reference for SFRC structural design.

Abstract: Based on the ultrasonic theory,the ultrasonic reflection method was suggested in the paper as nondestructive testing for steel bridge deck pavement layer bonding strength.According to five kinds of common interfaces in practical engineering,three types of related samples were made,and many detection tests were done by using the ZBL-520 non-metallic ultrasound instrument.By studying the waveform and acoustic parameters of each interface,it was found that the amplitude value could be used as the physical quantity to detect the bonding strength.Finally,the comparative experiments of loss and lossless interfaces were carried out on the same sample.The test results indicated that there were different degrees of differences in the amplitude of each interface,and the good bonding interface’s relative amplitude value was the smallest.The quality of bond layer of a part or the total of a bridge could be judged by the average value of a large number of measured point data.

Abstract: Based on Shuangji River aqueduct in the middle route of South-to-North Water Transfer Project,the structural type of large prestressed concrete rectangular aqueduct was examined.Plane structural mechanics method and 3d finite element method was used for calculation of reinforcement and concrete stress.By comparing the concrete stress,the groove construction method and the investment of different modes,it was concluded that slab structure of unribbed side wall and bottom plate was the appropriate structure for large-scale prestressed concrete aqueduct,and the structure of which had more balanced concrete stress,concrete cracking resistance,and was more convenient for mechanized construction.This paper provided a reference for the design and construction of large prestressed concrete rectangular aqueduct.

Abstract: A coupled structure-acoustic model of the car was built by hypermesh.Considering the excitation of the engine,prediction and analysis of the car inner noise were conducted.Using analysis of panel contribution,the key panels were found at peak frequencies and set these panel thickness as design variables.The response surfaces of mass and acoustic responses at peak frequencies were built by Latin Hypercube Sampling and error analysis was done based on Hammersley Sampling in order to select optimal response surfaces.Using selected response surfaces,mass as objectives,acoustic responses and vehicle stiffness as constraints,the optimization based on genetic algorithm was carried out.The effectiveness of the optimization result and improvement of optimization efficiency by response surface method were validated by the finite element model and the vehicle inner noise was improved.

Abstract: A series of hydroxyl and carboxyl functionalized ionic liquids had been designed and used as catalyst for the cycloaddition reaction of CO2 with epoxy compound.Meanwhile,the relationship between catalyst structure and catalytic performance was studied.The influence of the ratio of epoxy propane and catalyst,reaction pressure,reaction temperature,reaction time on the properties of catalyst had been inspected using cycloaddition of propylene oxide and CO2 as model reaction.The repeatability of catalyst was discussed.

Abstract: This paper used the Anyang Fifth People’s Hospital as the case of study.The experiments were implemented with the patterns of "2 pumping wells and 2 recharging wells" and "2 pumping wells and 4 recharging wells".The change of groundwater flow field and temperature field in the two patterns designed in ground temperature air-conditioning project operation were simulated by using a groundwater hydrodynamic model and a hydrothermal coupling mathematical model.Results indicated that the influence of ground temperature air-conditioning project on groundwater flow field and temperature field were very different in different pumping-recharging patterns.The increase of single-well recharge quantity enhanced the influence scope and range,making it easier to form heat transfixion.The way of groundwater transferring heat via groundwater also varied during different running periods.

Abstract: The smart express cabinets could solve terminal distribution problems to some extent.Based on the view of evolutionary game theory,an evolutionary game model of express enterprises and third-party smart express cabinet enterprises was established in this paper.Meanwhile,the dynamic evolutionary process and factors influencing the establishment of cooperative relationship were analyzed by this model.Finally,numerical cases were conducted with Matlab simulation software.The results showed that:(a)Only the excess income was more than the cost,probably the cooperation could be achieved.(b)Probability of this cooperation had a positive correlation with the excess income,while inversely with the cost.(c)There existed an optimal proportion of excess income correlation,which could maximize the probability of cooperation.(d) Reasonable income distribution mechanism and cost sharing mechanism were conducive to the establishment of the cooperative relationship.

Abstract: Based on steel making-continuous casting-hot rolling production process in iron and steel industry,the problem of scheduling n jobs in a multi-stage flexible flowshop with batching machines at some middle stage was studied.The batching production stage consisted of multiple serial batching machines in parallel,and the other stages contained discrete machines.Firstly,a mathematical model was formulated to minimize the total weighted completion time withthe consideration of job dynamic arrival,transportation time between the adjacent stages and machine setup time.Then,an improved adaptive genetic algorithm was developed for this NP-hard problem where the genetic parameters were associated with the iteration number and the fitness function values.Computational experiments tested a large number of random data for up to 150 jobs.The results show that the proposed algorithm could find the better solutions within a shorter period of time,as compared with the general genetic algorithm.The comparison with Lagrangian relaxation showed that the improved genetic algorithm performed better on solution quality for medium and large sized problems.

Abstract: In the running process of amusement equipment,equipment accidents and other problems have been arising-caused by high cycle fatigue of drive shaft.In this paper,the research on the fatigue simulation of the drive shaft inside the flying saucer amusement equipment is done using the finite element analysis theory and NCODE fatigue analysis software.Firstly,the amusement equipment is simplified and modeled according to the equipment working condition.Then,the ADAMS finite element analysis software is used to analyze the dynamics.As a result,the topological configuration of the multi-body system is presented and the load time history of the drive shaft is obtained.Next,by using NCODE fatigue analysis software,fatigue reliability analysis of the drive shaft is made,the nephograms of the fatigue results of the drive shaft and the fatigue life of each node are obtained,from which the location and fatigue life of the drive shaft easily destroyed are also confirmed.