Fault diagnosis is crucial for the stable and reliable operation of chemical processes. However, faced with complexity processes, conventional methods suffer from expertise feature extraction classifier design. They also lack extracting effective features raw data whose attributes are correlated coupled. This article proposes an enhanced naive Bayesian (ENBC) fault complex Assuming that all related, ENBC can utilize a joint probability density function (pdf) based on multivariate Gaussian...

Fault diagnosis that identifies the root of abnormal status is great importance to eliminate faults in complex chemical processes. Many data-driven fault models ignore different occur with varied frequencies plants, and they need a complete retraining process arrival new modes. In this article, novel incremental imbalance modified convolutional neural network proposed solve aforementioned issues. The method employs an extract valuable information from data, generate samples. After that,...

Fault diagnosis is of great importance to enhance the reliability and security complex chemical processes. However, many fault methods cannot extract correlation information among attributes in feature extraction procedure, resulting degradation performance. In this article, a novel framework proposed for First, framework, an enhanced kernel principal component analysis (eKPCA) obtain key features based on Hotelling's T <sup xmlns:mml="http://www.w3.org/1998/Math/MathML"...

This article proposes an adversarial reconstruction convolution neural network (ARCNN) for non-uniform illumination frontal face image recovery and recognition. The proposed ARCNN includes a discriminative network. authors employ GAN framework to learn the in manner. integrates gradient loss perceptual terms, which are able preserve detailed spatial structure information, into overall function constraint procedure. Experiments conducted on typical illumination-sensitive dataset, extended...

Relying on mechanical sound signals to carry out anomaly detection is a challenging task. Due the stability of production process complex industrial systems, there are very few or no abnormalities, and types failures also difficult describe in detail. In addition, characteristics machine itself will change with operating conditions, traditional models prone misjudge normal sounds as abnormal. We recommend that conditions similar situations be regarded domain shift between source target...

With the popularity of unattended stations, perimeter security technology has received close attention. Among many technologies, optical fiber sensor based method is most widely used one due to its passive, low-cost, long-life, strong anti-interference ability and environmental adaptability. This paper proposes an intrusion detection for by combing multiple features with improved double-threshold method. First, three types are extracted, including short-time energy, zero cross rate, wavelet...

Advice taking, which is a decision-making process formulated by decision-makers with the reference of others' suggestions, was investigated extensively in field behavioral decision making.To date, majority researchers have studied facilitating or hindering exchange advice situational factors, leaving plenty variables relatively unexplored.A small numbers studies that included individual differences, such as intelligence, conscientiousness, emotion, power, etc., found promising...

Ensemble semantic features are useful for acoustic scene classification. In this paper, we proposed a multi-scale fusion and channel weighted CNN framework. The framework consists of two stages: the feature weighting stages. stage extracts hierarchy maps using with simplified Xception architecture then integrates through top-down pathway. squeezes into descriptor transforms it set factors to reinforce importance each Experimental results on DCASE2018 classification subtask A B demonstrate...

Abstract Accurately and quickly obtaining the positions orientations of mechanical parts based on digital morphologies are key to achieving efficient accurate assembly parts. However, due poor robustness compactness in extracting morphology contours parts, accuracy obtained by using cannot meet requirements high-precision assembly. Therefore, this paper proposes a position orientation estimation method 3D contour registration. This extracts optimizes obtains an improved Iterative Closest...

A method measuring the depth of liquid surface oil well based on resonance principle air column is proposed in this paper. In method, casing pipe inspired by sending white noise to continuously, and then dynamic level can be calculated analysing frequency spectrum experimental signal. Aiming at problem that signal contains a lot noise, firstly, convolution window function with stronger filtering characteristics used improve ratio; Welch multi segment average power eliminate interference...

Abstract The high‐sulfur gas sweetening process is energy consuming. To analyze its efficiency effectively, and look for ways to improve this process, a novel energy‐efficiency evaluation method based on the yield‐energy‐selectivity coefficient methods was proposed. First, yield, energy, selectivity are selected as indicators establish model. Then, score of each indicator were determined by using analytical hierarchy efficacy method, respectively. Finally, proposed applied actual process....

We propose a deep reconstruction convolution neural network approach for illumination-robust face image recovery and recognition. Our learns an end-to-end mapping between non-uniform illumination uniform images. The architecture of the is functional divided into four stages, i.e., feature decomposition, multi-scale fusion, nonlinear reconstruction. illustrate that incorporating multiscale fusion stage in significantly remove unwanted information Besides, gradient constraint also added to...

Accidents in the chemical production process will have serious consequences, so fault diagnosis system is extremely important. To increase accuracy, it a method to expand size of neural network, but sensitivity test sample and training is. In this paper, using mixup-convolution network (CNN) model can extract more abundant information from time-varying features. Mixup uses neighborhood data better generalization ability overcome large-scale network. The problem remembering data. This...

Two dimensional linear discriminant analysis(2DLDA) provides a solution to the small sample size(S3) problem presented classical LDA. However, it takes each column of image matrix, which only contains partial information whole human face, as an input vector. In this paper, novel reassembling 2DLDA (R2DLDA) algorithm is proposed for face recognition. The new 2D sample, consisted sub-image original image, introduced. Then applied Experimental results on ORL database show that R2DLDA feasible...