Online Real-time Processing Research Articles

A SoC-FPGA based readout platform for the free-running AMBER data acquisition system

Owing to a change in the scope of physics research, the AMBER spectrometer at CERN is undergoing an update on its instrumentation and trigger strategy for the data acquisition system. One of the key updates is the adoption of a free-running and trigger-less operation, which differs from the predecessor of AMBER, COMPASS, by working in a continuous mode but in an event-based triggered way.In this article, we present a multichannel data-acquisition platform developed in line with the next generation of trigger-less and free-running data acquisition systems. The platform is based on a pre-existing Mezzanine Sampling ADC board for the analog-to-digital conversion and a Xilinx Zynq Ultrascale+ System on Module for online real-time data processing. We also present the evaluation of the system operating in continuous mode, taking data from a 25-element electromagnetic calorimeter prototype with a muon beam, during the first AMBER pilot run. The acquired data were used for noise and pulse shape studies required for the design of the algorithms for lossless compression and data feature extraction needed for trigger-less operation.

Online learning for data streams with bi-dynamic distributions

Data streams, as an important pattern of big data, require online real-time processing because instances arrive one by one and are fleeting. Existing online learning methods make distinctive assumptions, such as a fixed feature space, a varying feature space that follows specific patterns, and a fixed data distribution. However, data streams generated from real-world scenarios typically have both randomly changing feature spaces and data distributions, making existing methods inappropriate for practical applications. To fill this gap, this study proposes a novel Online Learning for Data Streams with Bi-dynamic Distributions (OLBD) algorithm. OLBD has a two-fold main idea: 1) it overcomes random changes in the feature space by building a mapping matrix to space transform and projects the original instances onto the global feature space; 2) it handles dynamic data distributions by constraining prior knowledge and transferring established mapping relationships to new distributions. To evaluate OLBD, we compared it with related state-of-the-art algorithms. First, we use 13 datasets to simulate three scenarios of dynamic feature space, namely trapezoidal, feature evolvable, and capricious data streams. Second, we simulated the data streams with dynamic data distributions using eight real and four generated datasets. We then conducted ablation studies on the parameter α. Finally, we analyzed data streams with bi-dynamic data distributions under different feature missing ratios and verified the generalization. The results show that OLBD significantly outperforms its rivals. Additionally, a practical case study on movie review classification was conducted to illustrate the effectiveness of OLBD in real-world scenarios.

A Bayesian Nuclide Identification Algorithm Based on the Energy Spectrum

A fast and reliable nuclide identification algorithm is important to the detection of radioactive materials, and the emergence of a sequential Bayesian algorithm provides a new direction of research on radioactive detection. This article proposes a Bayesian nuclide identification algorithm based on the energy spectrum and deploys the algorithm on a microprocessor for real-time online processing. The algorithm builds a probability model based on the natural background radiation and radioactive source radiation and a decision function based on sequence testing. The processor conducts a posteriori estimation on the collected energy spectrum at fixed time intervals and updates the decision function. A laboratory self-developed CeBr3 detector was used in testing. Four radioactive sources, namely 241Am, 133Ba, 137Cs, and 60Co, were used in experiments with fixed and moving radioactive sources. The experimental results show that the proposed algorithm can rapidly identify mixed and mobile radioactive sources.

First principles optimization of plutonium electrorefining

This work presents a means of controlling plutonium electrorefining at a maximum rate regardless of equipment setup through the derivation of power supply current and potential governing equations for normal and off-normal operations. The governing equations are demonstrated by electrorefining surrogate materials. A simple linear current sweeping method was used to determine the maximum electrorefining current for the surrogate system. This method can be used to develop autonomous process optimization, real-time online processing monitoring, and real-time process endpoint detection. Ultimately, this research provides the foundation to optimize the liquid metal electrorefining rate to decrease the time needed to the physical limit for the process.

Handheld portable neutron gamma discrimination spectrometer for environmental detection

A compact and integrated neutron-gamma discrimination spectrometer that could be utilized for personal dosimetry and emergency response is urgently required. In this study, based on a CLYC scintillator coupled to silicon photomultiplier(SiPM) arrays, we designed a stand-alone, hand-held neutron-gamma discrimination spectrometer. After the analog-to-digital converter (ADC) digitizes the analog signal from the detector, the data is transmitted to a FPGA processor for online real-time processing, and then transmitted to a microcontroller through the serial peripheral interface(SPI) protocol, and the current detection result is displayed in real-time on the touch screen. By using the pulse signal discrimination(PSD) method, the spectrometer can simultaneously measure the spectrum of neutron and gamma events, with a resolution of 8.8% (at 1274.5 keV) and excellent linearity. The figure of merit (FOM) is 2.32 with optimized PSD parameters. The experimental results show that the spectrometer integrates the upper computer and detector, with real-time performance, portability and excellent neutron-gamma resolution. Online operation is realized through the touch screen, which greatly facilitates the real-time detection of radiation in the field environment.

Real-time monitoring for manual operations with machine vision in smart manufacturing

Online real-time production process monitoring is the basis for intelligent manufacturing refinement management. This paper proposes a contactless monitoring framework with machine vision to realize real-time production/assembly process monitoring and repetition counting. In addition, a two-stage approach is presented to solve the issue of manual operation recognition and counting by pieces. In the first stage, a high-efficient method for object detection and classification, YOLOv4-tiny, is used to recognize manual operations/sub-operations and classify the sub-operations of the operations using monitoring video. And a Sliding Window Counter algorithm is utilized to find appropriate boundary points of different categories of the sub-operations for counting in the second stage. A dataset of repetitive manual operations is constructed for training this approach. The experiment results on the production/assembly line operations indicate that the approach can identify and count worker operations in real time.

Improving patient satisfaction and outpatient diagnostic center efficiency using novel online real-time scheduling

We develop a novel online real-time scheduling algorithm with applications for healthcare diagnostic centers to deal with walk-in patients based on a set of constraints on the sequence of tests and resources. The problem is especially significant at healthcare centers in developing and emerging nations, such as India, where appointment schedules do not work. Within this realistic context, our objective is to improve patient satisfaction by reducing waiting time and improve diagnostic center performance through better utilization of the constrained resources. We propose a Mixed Integer Linear Programming (MILP) formulation to represent diagnostic centers as a Flow and Open Shop, to capture the system dynamics of the Flexible Hybrid Shop Scheduling Problem. We then develop a novel Online Genetic Algorithm (OGA) capable of solving real life large scale problems, as Open Shop scheduling problems are NP-hard. The developed OGA is first validated for small instances against a theoretical lower bound and the MILP model using CPLEX solver for flow time and makespan. The OGA is then empirically validated with data collected from two diagnostic centers of different sizes and configurations. For both centers, the developed OGA shows significant improvement compared to the simulation model. This research offers an important contribution to both literature and practice as it is one of the first to model the patient scheduling problem as an online real-time process. Implementing the developed OGA would help diagnostic centers significantly improve time estimates, thus reducing actual patient time and improving the efficiency of the system. Most importantly, the OGA is generalizable beyond healthcare to a broad range of environments that share Hybrid Shop characteristics.

Illicit discharge detection in stormwater drains using an Arduino-based low-cost sensor network

Illicit discharges in urban stormwater drains are a major environmental concern that deteriorate downstream waterway health. Conventional detection methods such as stormwater drain visual inspection and dye testing have fundamental drawbacks and limitations which can prevent easy location and elimination of illegal discharges in a catchment. We deployed 22 novel low-cost level, temperature and conductivity sensors across an urban catchment in Melbourne for a year to monitor the distributed drainage network, thereby detecting likely illicit discharges ranging from a transitory flow with less than 10 minutes to persistent flows lasting longer than 20 hours. We discuss rapid deployment methods, real-time data collection and online processing. The ensemble analysis of all dry weather flow data across all sites indicates that: (i) large uncertainties are associated with discharge frequency, duration, and variation in water quality within industrial and residential land uses; (ii) most dry weather discharges are intermittent and transient flows which are difficult to detect and not simply due to cross-connections with the sewerage network; (iii) detectable diurnal discharge patterns can support mitigation efforts, including policies and regulatory measures (e.g., enforcement or education) to protect receiving waterways; and, (iv) that it is possible to cost effectively isolate sources of dry weather pollution using a distributed sensor network.

ESA-Stream: Efficient Self-Adaptive Online Data Stream Clustering

Many big data applications produce a massive amount of high-dimensional, real-time, and evolving streaming data. Clustering such data streams with both effectiveness and efficiency are critical for these applications. Although there are well-known data stream clustering algorithms that are based on the popular online-offline framework, these algorithms still face some major challenges. Several critical questions are still not answer satisfactorily: How to perform dimensionality reduction effectively and efficiently in the online dynamic environment? How to enable the clustering algorithm to achieve complete real-time online processing? How to make algorithm parameters learn in a self-supervised or self-adaptive manner to cope with high-speed evolving streams? In this paper, we focus on tackling these challenges by proposing a fully online data stream clustering algorithm (called ESA-Stream) that can learn parameters online dynamically in a self-adaptive manner, speedup dimensionality reduction, and cluster data streams effectively and efficiently in an online and dynamic environment. Experiments on a wide range of synthetic and real-world data streams show that ESA-Stream outperforms state-of-the-art baselines considerably in both effectiveness and efficiency.

An online log template extraction method based on hierarchical clustering

The raw log messages record extremely rich system, network, and application running dynamic information that is a good data source for abnormal detection. Log template extraction is an important prerequisite for log sequence anomaly detection. The problems of the existing log template extraction methods are mostly offline, and the few online methods have insufficient F1-score in multi-source log data. In view of the shortcomings of the existing methods, an online log template extraction method called LogOHC is proposed. Firstly, the raw log messages are preprocessed, and the word distributed representation (word2vec) is used to vectorize the log messages online. Then, the online hierarchical clustering algorithm is applied, and finally, log templates are generated. The experimental analysis shows that LogOHC has a higher F1-score than the existing log template extraction methods, is suitable for multi-source log data sets, and has a shorter single-step execution time, which can meet the requirements of online real-time processing.

Watching and Safeguarding Your 3D Printer

The increasing adoption of 3D printing in many safety and mission critical applications exposes 3D printers to a variety of cyber attacks that may result in catastrophic consequences if the printing process is compromised. For example, the mechanical properties (e.g., physical strength, thermal resistance, dimensional stability) of 3D printed objects could be significantly affected and degraded if a simple printing setting is maliciously changed. To address this challenge, this study proposes a model-free real-time online process monitoring approach that is capable of detecting and defending against the cyber-physical attacks on the firmwares of 3D printers. Specifically, we explore the potential attacks and consequences of four key printing attributes (including infill path, printing speed, layer thickness, and fan speed) and then formulate the attack models. Based on the intrinsic relation between the printing attributes and the physical observations, our defense model is established by systematically analyzing the multi-faceted, real-time measurement collected from the accelerometer, magnetometer and camera. The Kalman filter and Canny filter are used to map and estimate three aforementioned critical toolpath information that might affect the printing quality. Mel-frequency Cepstrum Coefficients are used to extract features for fan speed estimation. Experimental results show that, for a complex 3D printed design, our method can achieve 4% Hausdorff distance compared with the model dimension for infill path estimate, 6.07% Mean Absolute Percentage Error (MAPE) for speed estimate, 9.57% MAPE for layer thickness estimate, and 96.8% accuracy for fan speed identification. Our study demonstrates that, this new approach can effectively defend against the cyber-physical attacks on 3D printers and 3D printing process.

Exploring Implicit Feedback for Open Domain Conversation Generation

User feedback can be an effective indicator to the success of the human-robot conversation. However, to avoid to interrupt the online real-time conversation process, explicit feedback is usually gained at the end of a conversation. Alternatively, users' responses usually contain their implicit feedback, such as stance, sentiment, emotion, etc., towards the conversation content or the interlocutors. Therefore, exploring the implicit feedback is a natural way to optimize the conversation generation process. In this paper, we propose a novel reward function which explores the implicit feedback to optimize the future reward of a reinforcement learning based neural conversation model. A simulation strategy is applied to explore the state-action space in training and test. Experimental results show that the proposed approach outperforms the Seq2Seq model and the state-of-the-art reinforcement learning model for conversation generation on automatic and human evaluations on the OpenSubtitles and Twitter datasets.

Rapid resolving algorithm of gas chromatographic overlapping peaks based on the asymptotic expansion of integration

A novel algorithm, called asymptotic expansion of integration, is suggested to resolve gas chromatographic overlapping peaks. There are three steps for the algorithm. First, a valley peak or a shoulder peak is separated into two domains, and an integral equation on a subdivision and an algebraic equation on the overlapping peak domain are listed. Secondly, areas needed in two equations, are computed by a numerical integral method, then the integral equation is expended to an algebraic equation by the asymptotic formula of integration. At last, combing two equations with constraint equations of peak heights, we got a nonlinear algebraic set. The equation set can be solved rapidly by Gauss-Seidel iteration, and the maximum number of iterations is not more than 20 times. The simulation and experimental results showed that height and area errors of resolving peaks are quite small, the maximum error of area is less than 6.44%, and that of the height is about 6.80%. Because of the high accuracy and computational efficiency, the algorithm can be used in decomposition of gas chromatographic overlapping peaks and online real-time processing of general chromatographic overlapping peaks.

A Noise Reduction Method for MEMS Gyroscope Based on Direct Modeling and Kalman Filter

This paper focus on the topic of reducing the random noise in the output data of MEMS gyroscope and improving the accuracy of MEMS gyroscope. Aiming at the problems that the previous methods are not suitable for dynamic data and difficult to realize real-time online processing, a direct modeling method is proposed to establish the output data model of MEMS gyroscope. Based on this model, the kalman filter is designed to process the output data. The experimental results show that the proposed method has good performance on both static data and dynamic data.

Profit Per Hour as a Target Process Control Parameter for Manufacturing Systems Enabled by Big Data Analytics and Industry 4.0 Infrastructure

The rise of Industry 4.0 and in particular Big Data analytics of production parameters offers exciting new ways for optimization. The majority of factories in process industries currently aim for example, either for output maximization, yield increase, or cost reduction. The availability of real-time data and online processing capability with advanced algorithms enables a profit per hour operational management approach. Profit per hour as a target control metric allows running factories at the optimal available operating point taking all revenue and cost drivers into account.This paper describes the suitability of profit per hour as a target process control parameter for production in process industries. The authors explain how this management approach helps to make better operational decisions, trading off yield, energy, throughput, among other factors, and the resulting cumulative benefits. They also lay out how Big Data and advanced algorithms are the key enabler to this new approach, as well as a standardized methodology for implementation. With profit per hour an agile control approach is presented which aims to optimize the performance of industrial manufacturing systems in a world of ever increasing volatility.

Review of intelligent microblog short text processing

This article summarizes the research methods for processing short texts in the microblog, one of the most popular social media sites, analyzes the bottleneck problems and challenging issues, and discusses the possible direction and strategies for solving the above problems. Comprehensive literature review on short text processing has also been conducted by the way of case tracking and case studies. Text categorization and text clustering are the main research methods for processing short texts in microblogs, and topic detection, sentiment analysis, smart healthcare, business intelligence and location services are hot topics for the application of these methods. We also found some challenges for processing microblog short texts, including semantic processing, noise reduction, clarification of fuzzy themes and handling big data. The real-time online processing of big data is expected to be the main research direction in the future.

Recurrence CFD – A novel approach to simulate multiphase flows with strongly separated time scales

Classical Computational Fluid Dynamics (CFD) of long-time processes with strongly separated time scales is computationally extremely demanding if not impossible. Consequently, the state-of-the-art description of such systems is not capable of real-time simulations or online process monitoring.In order to bridge this gap, we propose a new method suitable to decouple slow from fast degrees of freedom in many cases. Based on the recurrence statistics of unsteady flow fields, we deduce a recurrence process which enables the generic representation of pseudo-periodic motion at high spatial and temporal resolution.Based on these fields, passive scalars can be traced by recurrence CFD. While a first, Eulerian Model A solves a passive transport equation in a classical implicit finite-volume environment, a second, Lagrangian Model B propagates fluid particles obeying a stochastic differential equation explicitly.Finally, this new concept is tested by two multiphase processes – a lab scale oscillating bubble column and an industrial scale steelmaking converter. Results of tracer distribution obtained by recurrence CFD are in very good agreement with full CFD simulations, while computational times are dramatically reduced. Actually, recurrence CFD is a promising candidate for online simulations of passive transport processes at full CFD resolution, which opens the door towards improved process monitoring.

Cloud based Real-time Multi-robot Collision Avoidance for Swarm Robotics

Nowadays, Cloud Computing has brought many new and efficient approaches for computation-intensive application areas. One typical area is Cloud based real-time device control system, such as the IoT Cloud Platform. This kind of platform shifts computation load from devices to the Cloud and provides powerful processing capabilities to a simple device. In Swarm robotics, robots are supposed to be small, energy efficient and low-cost, but still smart enough to carry out individual and swarm intelligence. These two goals are normally contradictory to each other. Besides, in real world robot control, real-time on-line data processing is required, but most of the current Cloud Robotic Systems are focusing on off-line batch processing. However, Cloud based real-time device control system may provide a way that leads this research area out of its dilemma. This paper explores the availability of Cloud based real-time control of massive complex robots by implementing a relatively complicated but better performed local collision avoidance algorithm. The Cloud based application and corresponding Cloud driver, which connects the robot and the Cloud, are developed and deployed in Cloud environment. Simulation tests are carried out and the results show that, when the number of robots increases, by simply scaling computation resources for the application, the algorithm can still maintain the preset control frequency. Such characteristics verify that the Cloud Computing environment is a new platform for studying massive complex robots in swarm robotics.

Predicting trace organic compound attenuation with spectroscopic parameters in powdered activated carbon processes

The removal of trace organic compounds (TOrCs) is of growing interest in water research and society. Powdered activated carbon (PAC) has been proven to be an effective method of removal for TOrCs in water, with the degree of effectiveness depending on dosage, contact time, and activated carbon type. In this study, the attenuation of TOrCs in three different secondary wastewater effluents using four PAC materials was studied in order to elucidate the effectiveness and efficacy of PAC for TOrC removal. With the notable exception of hydrochlorothiazide, all 14 TOrC indicators tested in this study exhibited a positive correlation of removal rate with their log Dow values, demonstrating that the main adsorption mechanism was hydrophobic interaction. As a predictive model, the modified Chick-Watson model, often used for the prediction of microorganism inactivation by disinfectants, was applied. The applied model exhibited good predictive power for TOrC attenuation by PAC in wastewater. In addition, surrogate models based upon spectroscopic measurements including UV absorbance at 254 nm and total fluorescence were applied to predict TOrC removal by PAC. The surrogate model was found to provide an excellent prediction of TOrC attenuation for all combinations of water quality and PAC type included in this study. The success of spectrometric parameters as surrogates in predicting TOrC attenuation by PAC are particularly useful because of their potential application in real-time on-line sensor monitoring and process control at full-scale water treatment plants, which could lead to significantly reduced operator response times and PAC operational optimization.

Recognition of one Class of Quadrics from 3D Point Clouds

Within cyber physical production systems 3D vision as a source of information from real-world provides enormous possibilities. While the hardware of contemporary 3D scanners is characterized by high speed along with high resolution and accuracy, there is a lack of real-time on-line data processing algorithms that would give certain elements of intelligence to the sensory system. Critical elements of data processing software are efficient, real-time applicable methods for fully automatic recognition of high level geometric primitives from point cloud (surface segmentation and fitting). This paper presents a method for recognition of one class of quadrics from 3D point clouds, in particular for recognition of cylinders, elliptical cylinders and ellipsoids. The method is based on the properties of scatter matrix during direct least squares fitting of ellipsoids. Presented recognition procedure can be employed for segmentation of regions with G1 or higher continuity, and this is its comparative advantage to similar methods. The applicability of the method is illustrated and experimentally verified using two case studies. First case study refers to a synthesized, and the second to a real-world scanned point cloud.