In order to solve the problem of physical constraints of the DC(Direct-Current), an MPC (Model
Predictive Control) controller tracks the speed of brush DC motor. Based on the established motor model, the
MPC controller is designed to track the expected speed if the brush DC motor, the formed QP(Quadratic
Programming) problem is solved by path-following method. A real-time simulation platform is built to verify the
controller. The result shows that the MPC controller has good controlling performance.

In order to improve the success probability of number estimation of signals in the presence of color
noise, an Otsu method based on pseudocovariance matrix is proposed. The pseudocovariance matrix composed of
delay correlation of array output signals is immune to white Gaussian noise and colored Gaussian noise under
certain conditions. The array’s effective aperture is improved by using the time correlation between arrays, and
then the differences between signal singular values and noise singular values, which are obtained from the
singular value decomposition of pseudocovariance matrix, are also improved. And by utilizing Otsu method to
classify the signal singular values and noise singular values, the probability of successful estimation of source
number is further improved.

 Because PMLSM ( Permanent Magnet Linear Synchronous Motor) does not have intermediate
transmission links, any uncertainty disturbances will directly affect the performance of the control system and
conventional ADRC (Auto Disturbance Rejection Control) parameters many is not conducive to tuning problems,
we proposed a improved invasive weed optimization PMLSM two order ADRC. We use chaos reverse learning
initialization method and the Cauchy distribution of spatial distribution improved invasive weed algorithm to
optimize the parameters of ADRC. And after optimization of ADRC, the control performance has improved
significantly. Through simulation and comparison, the improved auto disturbance rejection controller has fast
response speed, small steady-state error decreases by 2%, and does not have overshoot, and has good robustness
to load disturbance.

Feature extraction is a basis, a vital step and a major issue in facial expression recognition. To
ensure that the extracted features can be more comprehensive characterization of a certain kind of expression,
we present a feature extraction method based on fused geometry and local texture features. Geometric features
are obtained from the feature points marked by AAM (Active Appearance Model) algorithm, texture feature
extraction is based on LBP (Local Binary Pattern) algorithm, the dimension of fusion expression features is
reduced by LLE ( Locally Linear Embedding) algorithm. Finally, a multi-class SVM ( Support Vector
Machine) is used for facial expression classification. Our method is deployed on the JAFFE and Yale data
sets, the results show a recognition accuracy of 98. 57% and 91. 67% respectively, which prove the
effectiveness of our proposed method.

 In order to solve the problem of the stability analysis of the linear descriptor fractional-order systems,
the continuous frequency distributed equivalent model and the indirect Lyapunov approach are applied in the
research of the linear descriptor fractional-order systems, an PD(Proportional-Derivative) controller is proposed
to normalize the descriptor fractional order system. A novel sufficient condition for asymptotic stability of the
descriptor fractional-order system with the order α(0<α<1) is presented. LMI(Linear Matrix Inequalities)
techniques and matrix’s SVD (Singular Value Decomposition) are used to obtain the results. A numerical
example is provided to demonstrate the effectiveness of the proposed methods.

 Aiming at the reservoir lithology discrimination based analysis data of coring well and the logging
process data, we proposed and established a learning vector quantization process neural network model (LVQ-
PNN). The model by increasing the output layer, expanded the depth of the self-organization process neural
network, using the combination algorithm strategies with unsupervised competition and supervised learning,
to improve the adaptive extraction of multidimensional signal feature and self-organization comprehensive ability,
The experimental result shows that it has better recognition ability and comprehensive lithology. The recognition
rate of lithology discrimination is 84. 7%.

In order to break through the limitation of traditional study on behavior of flow and mass transfer by
material objects, CFD ( Computational Fluid Dynamic) simulations of PABR ( Periodic Anaerobic Baffled
Reactor) on RTD(Residence Time Distribution) was carried out in this paper, supporting implementing schemes
to economize manpower and material resources in configuration improvement, simulation of operating mode, risk
monitoring and forecast. A three-dimensions modeling of the tap water body in the PABR modelled by SOLID
WORKS was meshed by COMSOL. Then, RTD simulating test was performed by unidirectional flow layer
module, thin mass transfer module and transient solver, based on the boundary conditions of lab-scale
experimental material object. Normalized integration of the normalized curves of both the material object and the
simulating was fitted and calculated by Matlab, and the similarity of the two curves was acceptable. Finally, the
behavior of tracer in the flow and diffusion procedure was visualized by post-processing with COMSOL, and
human-computer interaction was achieved.

Motion data retrieval is used to get the proper data from the existing motion database for motion editing
and reusing, the hierarchical retrieval based on posture feature coding improves the retrieval speed and quality.
We divide major bone into nine segments, and the angles of these bone segments can response geometric relations
of motion posture, can used as feature vectors, encode posture features, create database index based on the
motion features, in the retrieval process, carry out the similarity comparison from motion feature, data index and
motion data value. The experimental results shows, in the motion database that contains walk, run and dance
postures, compared with the other method, the method has high retrieval speed and satisfactory retrieval results,
and realize the effectively retrieve of the logical similarity motion data.

The nonlinear relationship between forest fuel moisture and meteorological factors is an important
theoretical basis for the occurrence of forest fire and the prevention of fire. The moisture content of surface dead
fuel in three typical stands in Harbin urban city forestry demonstration base is taken as the object of study.
During the fall from September 2014 to November, the fixed two days are selected every week. The results
showed that the moisture content and meteorological factors of the combustible species in the three forest types of
Quercus mongolica, Juglans mandshurica, Betula platyphylla were studied from 8:00 to 18:00, and the dead
leaves, the 10 hr dead branches and the 1hr dead branches were selected to analyze nonlinear relationship
between moisture content and meteorological factors of materials, EXCEL 2013 and SPSS19. 0 are used on the
data for follow-up processing and correlation analysis. The results showed that: it was found that moisture of fuel
in the three forest types ranged from 14% to 67%; In the nonlinear model It is found that the correlation between
one hour ago and the current time is the highest that reaches 0. 89. In the nonlinear exponential model, the
model accuracy of Quercus mongolica and White birch was less than 15%. The accuracy of the model is better.

In order to improve the detection methods of oil pipeline leakage, this article briefly introduces some
commonly used methods of oil pipeline leakage detection, such as direct detection methods, negative pressure
wave detection method and neural network based detection method and so on. These methods have their
advantages and disadvantages in application. However, with the improvement of oil pipeline leak detection
requirements, these methods cannot meet people’s requirements, still need to be further improved.
Subsequently, we introduce the deep learning. Deep learning is the development on the basis of neural network
and its application in many aspects makes up the deficiency of the application. Among them, deep learning has
been successful in image and speech recognition applications. These facts provide some theoretical support for the
subsequent application of deep learning in oil pipeline leak detection.

A new relation extraction method is proposed to solve the problem of relation extraction from open
Chinese free texts. In order to alleviate the difficult problem of relation triples extraction, a method based on the
relationship between attribute and concept instance triples is proposed, a large number of instances of concept
and relation triples includes explicit relation triples and contains an implicit relation triples. The relationship
triple construction contains noise and error, in view of the relationship between the ternary group is used contains
noise and wrong question, Adaboost based iterative algorithm of collaborative training methods is used to
strengthen the relationship between extraction model. Experiment is carried out on the text of the encyclopedia
entries in the field of university, and the experimental results show that the method can obtain better
performance.

Traditional emotional evaluate methods are mainly based on changes in facial expressions and body
movements. To solve the subjective problems existing in these methods, we use positive and negative emotional
EEG (Electroencephalagrams) signal as the research object, anddesign the picture watching experiments with
the help of the CAPS (Chinese Affective Picture System) of Chinese Academy of Sciences, and then analyze the
final data using machine learning algorithm of SVM (Support Vector Machine). Moreover, according to the
actual situation of the experiment, this paper presents a new method that dividing 12 consecutive data into one set
and using the mean of each set into emotional evaluating. The results show that this new method obtainsa better
recognition accuracy of 73. 33%. This research not only provides a method that reduces the error caused by
emotion generating delay, but also provides a reference for the analysis and optimization of brain waves.

In order to solve the problem that the existing scheduling task scheduler Hadoop can not depend on
the degree of urgency task allocation of resources. We mainly research the resource scheduling mechanism in
YARN (Yet Another Resource Negotiator), improved the original capacity scheduler and proposes abased on
priority weighted Hadoop YARN scheduling algorithm-weight scheduler based on priority. Firstly, setting queue
priority for leaf queue, combining with the queue resource utilization and queue priority to select queue;
Secondly, the initial of applications re-set its priorities size, by waitting time to judge whether to update the
weight, according to the weight to sort the applications in the selected queue. The experiment shows that
comparing with the existing scheduling algorithm, this algorithm has the advantage on the execution time by
141 s and waiting time by 34. 5% decreasing, ensure the user to perform a task relatively fair, improve the
overall satisfaction of users.

 In order to solve the problems of Jilin province science and technology information service platform,
where the computer room management is overloaded, equipment maintenance is difficult and data can not be
collected, etc. We design a computer room management duty system. The design of this system can solve the
room management difficulties very well, achieve the electronic, remote control and networking of the computer
room management. It has certain practical application value.

 In order to establish and improve the special database of oil sands spectrum, we introduce the design
and development of special database of oil sands spectrum, which uses C# Language and SQL Server 2008
database to conduct the data administration and provide the data description. 90 group of oil sands spectrum data
in the wild of Qinghai province are used for database testing, tests prove that: the parameters of the characteristic
of spectral absorption including position, depth, width, slope, area, and symmetry of absorption wave length can
be obtained. The unknown measured spectral curue and well-known spectrum library curve can be matched with
each other. It meets the requirements of the database design.

 In order to better improve student dormitory fire and burglar alarm ability, we provide students with a
safe and convenient living environment. We design the fire protection and anti-theft system which based on
WeChat public platform using the wireless sensor network technology and background server technology. Users
can use the system to check indoor conditions in real-time, which achieves remote monitoring of student
dormitories. The system is low-cost, flexible and user-friendly. It has gained a good user experience after
actual use.

 In order to solve the instability problem of the bus communication system for engineering vehicles, a
design structure of CAN(Controller Area Network) bus software system based on engineering vehicle safety is
proposed. The operating environment of the engineering vehicle bus system is analyzed, its interference sources
and the current anti-jamming measures are summarized. Focusing on the system fault tolerance and real-time
problem, the engineering vehicle safety CAN bus software system is designed. The specific functions of various
types of nodes and the realization of the program. The software design principle adopts the modular and
standardized system design method, reduces the dependence of the system software to the specific manufacturer,
improves the safety of the vehicle and enhances the robustness and real-time of the vehicle bus system.