Quantification of human behavior levels by extending Rasmussen's SRK model and the effects of time pressure and training on the levels switching.

Human factor-related accidents account for an increasing portion of the total accidents through the advancing level of system automation. Human reliability becomes the key issue in human-machine systems especially for safety-relevant tasks and operations. Rasmussen's SRK (skill-rule-knowledge) model is well known in the field of human factors. Likewise, it is well known that skill-based behaviors have the highest human reliability, while knowledge-based behaviors are associated with the lowest reliability scores. Although numerous studies exist on human error probability (HEP), correspondingly typically attributed directly or indirectly to these three levels of behavior, a coherent, consistent representation, especially using data sources, has not been available. In this contribution, the quantification of human behavior levels with Rasmussen's SRK model is given based on three databases for the first time. Effects of time pressure and training on human reliability switching are also analyzed based on related publications. To determine the HEP of these three levels, three databases, technique for human error rate prediction (THERP), Savannah river site human reliability analysis (SRS-HRA) and nuclear action reliability assessment (NARA), from human reliability analysis (HRA) methods are used. The procedure contains identifying the tasks including the operator involved and the assumptions the analysts made and classifying the tasks into suitable cognitive behavior mode (CBM). In this case, the relationship between SRK levels and HEP is mapped. The effects of the two in automation context very relevant performance shaping factors (PSFs), time pressure and training/knowledge degradation, on human behavior levels switching are analyzed and the explanations of the SRK switching are presented. In this case, a more general structure is established to illustrate the dynamic behavior of levels switching with six directions under different conditions. From the results we conclude that skill, rule, and knowledge behavior levels are continuous in terms of HEP and therefore allow a new inside into this key aspect of human factor quantification. Based on this analysis the consequences of daily automation in the context of autonomous transport systems in combination with human qualification and reliability degrading are from this specific and in the current automation discussion very intensively discussed. The presented discussion linking SRK levels and HEP gives a new perspective on the foreseeable consequences of further automation in application areas with increasing automation of everyday tasks (like using a highly automated vehicle).

Automated Measurement and Control of Germination Paper Water Content.

Germination paper (GP) is used as a growth substrate in plant development studies. Current studies bear two limitations: (1) The actual GP water content and variations in GP water content are neglected. (2) Existing irrigation methods either maintain the GP water content at fully sufficient or at a constant deficit. Variation of the intensity of water deficit over time for plants grown on GP is not directly achievable using these methods. In this contribution, a new measurement and control approach was presented. As a first step, a more precise measurement of water content was realized by employing the discharging process of capacitors to determine the electrical resistance of GP, which is related to the water content. A Kalman filter using an evapotranspiration model in combination with experimental data was used to refine the measurements, serving as the input for a model predictive controller (MPC). The MPC was used to improve the dynamics of the irrigation amount to more precisely achieve the required water content for regulated water uptake in plant studies. This is important in studies involving deficit irrigation. The novel method described was capable of increasing the accuracy of GP water content control. As a first step, the measurement system achieved an improved accuracy of 0.22 g/g. The application of a MPC for water content control based on the improved measurement results in an overall control accuracy was 0.09 g/g. This method offers a new approach, allowing the use of GP for studies with varying water content. This addressed the limitations of existing plant growth studies and allowed the prospection of dependencies between dynamic water deficit and plant development using GP as a growth substrate for research studies.

Explorative Frequency Analysis of Leaf Temperature Behavior of Maize (Zea mays subsp. mays) at Water Deficit.

In this study, different standard frequency analysis (FA) methods are applied to measured leaf temperature data of maize plants (developmental stages EC13-15). These FA methods are used to identify specific behaviors, regularities, and sudden changes in frequencies/amplitudes of data, e.g., in control engineering. The thorough application of different FA methods in plant studies is novel. The aim of this paper is to analyze features of the measured data and to explore the explanatory power of different methods for the detection of plant dynamic behavioral changes. The basic assumption is an expected relation between plant water stress and resulting changes in leaf temperature oscillations caused by stress-induced changes in stomatal behavior. Therefore, an irrigation experiment (laboratory; controlled environmental conditions) was implemented to compare leaf temperature behavior of stressed and unstressed plants. Leaf temperature time series are processed and the results are compared as functions of time showing the behavioral changes in terms of the different methods applied. The analysis of results is explained; conclusions, which can be made based on different methods, are given. The study confirms the applicability of FA methods and provides new insights into leaf temperature behavioral patterns. Results are discussed regarding the hypothesized incipience of leaf temperature oscillations due to water stress.

Bearing Fault Diagnosis Method Based on Deep Convolutional Neural Network and Random Forest Ensemble Learning.

Recently, research on data-driven bearing fault diagnosis methods has attracted increasing attention due to the availability of massive condition monitoring data. However, most existing methods still have difficulties in learning representative features from the raw data. In addition, they assume that the feature distribution of training data in source domain is the same as that of testing data in target domain, which is invalid in many real-world bearing fault diagnosis problems. Since deep learning has the automatic feature extraction ability and ensemble learning can improve the accuracy and generalization performance of classifiers, this paper proposes a novel bearing fault diagnosis method based on deep convolutional neural network (CNN) and random forest (RF) ensemble learning. Firstly, time domain vibration signals are converted into two dimensional (2D) gray-scale images containing abundant fault information by continuous wavelet transform (CWT). Secondly, a CNN model based on LeNet-5 is built to automatically extract multi-level features that are sensitive to the detection of faults from the images. Finally, the multi-level features containing both local and global information are utilized to diagnose bearing faults by the ensemble of multiple RF classifiers. In particular, low-level features containing local characteristics and accurate details in the hidden layers are combined to improve the diagnostic performance. The effectiveness of the proposed method is validated by two sets of bearing data collected from reliance electric motor and rolling mill, respectively. The experimental results indicate that the proposed method achieves high accuracy in bearing fault diagnosis under complex operational conditions and is superior to traditional methods and standard deep learning methods.

Development of A Low-Cost FPGA-Based Measurement System for Real-Time Processing of Acoustic Emission Data: Proof of Concept Using Control of Pulsed Laser Ablation in Liquids.

Today, the demand for continuous monitoring of valuable or safety critical equipment is increasing in many industrial applications due to safety and economical requirements. Therefore, reliable in-situ measurement techniques are required for instance in Structural Health Monitoring (SHM) as well as process monitoring and control. Here, current challenges are related to the processing of sensor data with a high data rate and low latency. In particular, measurement and analyses of Acoustic Emission (AE) are widely used for passive, in-situ inspection. Advantages of AE are related to its sensitivity to different micro-mechanical mechanisms on the material level. However, online processing of AE waveforms is computationally demanding. The related equipment is typically bulky, expensive, and not well suited for permanent installation. The contribution of this paper is the development of a Field Programmable Gate Array (FPGA)-based measurement system using ZedBoard devlopment kit with Zynq-7000 system on chip for embedded implementation of suitable online processing algorithms. This platform comprises a dual-core Advanced Reduced Instruction Set Computer Machine (ARM) architecture running a Linux operating system and FPGA fabric. A FPGA-based hardware implementation of the discrete wavelet transform is realized to accelerate processing the AE measurements. Key features of the system are low cost, small form factor, and low energy consumption, which makes it suitable to serve as field-deployed measurement and control device. For verification of the functionality, a novel automatically realized adjustment of the working distance during pulsed laser ablation in liquids is established as an example. A sample rate of 5 MHz is achieved at 16 bit resolution.