Sensor Model Research Articles

Interval type-2 fuzzy stochastic configuration networks for soft sensor modeling of industrial processes

Soft sensors have been widely applied to predict key variables that are difficult to measure for industrial process modeling. In this paper, a novel randomized interval type-2 fuzzy neural network with parallel learning, called IT2F-PSCN, is presented for soft sensor modeling of industrial processes. It trains the upper and lower bounds of the interval type-2 fuzzy logic system separately to facilitate type reduction, thereby integrating the fuzzy logic system with stochastic configuration networks. To achieve the appropriate structure and parameters of the model, we develop a two-phase training scheme. In the first phase, a sparse rule interpolation method with stochastic configuration is applied to generate new fuzzy rules. In the second phase, the hidden layer is constructed through parallel stochastic configuration to enhance the nonlinear representational capacity. The validity of IT2F-PSCN is confirmed by a series of experiments, including four benchmark data modelings, simulation on the Tennessee Eastman process, and soft sensor modeling for the slurry grade of the first rougher in a zinc flotation process. The experimental results indicate that the proposed IT2F-PSCN performs favorably compared with other methods.

Harnessing NOx emission management: A virtual sensor model for natural gas power generation engines with active pre-chamber

This paper presents a NOx Sensor machine learning model for a highly efficient 2MW power generation lean burn gas engine with an active pre-chamber. As gas power generation continues to evolve, engine efficiency and the accurate measurement of NOx emissions has taken on greater importance. However, the working life of existing NOx sensors pales in comparison to the lifecycle of an engine. This paper studies a series of machine learning NOx sensor models and carries out a comparative analysis. We introduce a predictive analytics model used for monitoring the emissions of active pre-chamber gas engines. This real deployment aims to detect the deviations in emissions across an entire fleet of gas engines, even when no physical sensors are installed in the engines. Finally, a model has been implemented that allows the prediction of NOx values, in engines that did not have them initially, with a 5–10 % precision. For example, with NOx at 120 ppm, the virtual sensor provides values close to +-10 ppm. It has been observed that such features as the setpoint value of the active pre-chamber gas pressure, in a lean burn gas engine with active pre-chamber, are among those parameters that have the greatest impact on the model estimation.

Research on semi-supervised soft sensor modeling method for sulfur recovery unit based on ISSA-VMD-ESN

Sulfur recovery unit (SRU) in the refining industry can effectively reduce sulfide emissions and achieve efficient use of resources. Predicting sulfide emissions with soft sensors can provide the basis for process optimization and control, to improve SRU’s stability and efficiency. However, due to the dynamicity and nonlinearity of the sulfur recovery process, existing soft sensor models have poor prediction accuracy. Therefore, this paper proposes a semi-supervised soft sensor modeling method based on variational mode decomposition (VMD), echo state network (ESN) and improved sparrow search algorithm (ISSA). Firstly, Reconstructing SRU dataset using semi-supervised fusion method to reduce the impact of dynamicity on prediction results. Secondly, through the ISSA to optimize the VMD parameters and using the optimized VMD to decompose the output sequences into multiple components, to reduce the nonlinearity of sequences. Then, the soft sensor modeling of each component using ESN. Additionally, the ESN parameters are optimized by ISSA to prevent the effect of improper parameter settings on the model’s prediction performance. Afterwards, the predicted values of each component are superimposed to obtain the final prediction. Finally, simulation experiments are conducted based on the SRU dataset of an oil refinery to verify the effectiveness and generalization of the proposed method.

Introducing the double validation metric for radar sensor models

In automated vehicles, environment perception is performed by various sensor types, such as cameras, radars, lidars, and ultrasonics. Simulation models of these sensors, as required in virtual validation methods, are available in various degrees of detail. However, proving the validity of such models is a subject of research. New metrics and methods for credibility assessment of simulation are needed to standardize the validation process in the future. The so-called double validation metric (DVM) has shown advantages and allows an intuitive interpretability of the validation results. The DVM has so far only been applied to lidar sensor models. In this paper, an extension to the DVM is introduced, which is called the DVM Map. A static measurement scenario is conducted in reality and transferred into simulation. The novel method is demonstrated on the obtained real and simulated radar sensor data. In this simple scenario special focus is put on the position accuracy of GNSS reference sensors. Therefore, their impact on the result of sensor model validation is discussed. The paper shows that the method provides a more detailed and accurate validation in comparison to the state of the art of a radar simulation, revealing previously undetected simulation errors. Errors due to the environment model, signal propagation, and signal processing are separated and satellite imagery is used for intuitive visualization of the results. This method is a complementary tool to existing validation techniques to improve the interpretability and judging the trustworthiness of radar simulations.

An enhanced fireworks algorithm and its application in fault detection of the displacement sensor

Regarding the fault detection problem of the displacement sensor, an enhanced fireworks algorithm with information crossover and conversion factor (EFWA-IC) is proposed in this paper. In EFWA-IC, an information crossover strategy is proposed to maintain population diversity. This strategy is based on the predatory behavior of carnivores in nature. In addition, a conversion factor is set to control whether the explosion operator is executed or not to provide a more reasonable search. To fully evaluate the performance of EFWA-IC, a range of tests are carried out based on the CEC2017 and 23 classical test functions. The results show that the performance of EFWA-IC is better than other state-of-the-art optimization algorithms in terms of solution accuracy, convergence speed, and stability. Finally, EFWA-IC is utilized to optimize the particle filter (PF) to establish a fault detection model of displacement sensor in the continuous casting mold. The simulation experiment result of field data manifests that EFWA–IC–PF can accurately detect faults in the displacement sensor. For bias faults, the RMSE of the EFWA–IC–PF model is 0.14468, and the false alarm rate (FAR) and missed detection rate (MDR) are 1 % and 0.5 %, respectively. For stuck faults, the RMSE is 1.00148, and the FAR and MDR are 0.88 % and 1 %, respectively.

Adversarial Robustness Enhancement for Deep Learning-Based Soft Sensors: An Adversarial Training Strategy Using Historical Gradients and Domain Adaptation.

Despite their high prediction accuracy, deep learning-based soft sensor (DLSS) models face challenges related to adversarial robustness against malicious adversarial attacks, which hinder their widespread deployment and safe application. Although adversarial training is the primary method for enhancing adversarial robustness, existing adversarial-training-based defense methods often struggle with accurately estimating transfer gradients and avoiding adversarial robust overfitting. To address these issues, we propose a novel adversarial training approach, namely domain-adaptive adversarial training (DAAT). DAAT comprises two stages: historical gradient-based adversarial attack (HGAA) and domain-adaptive training. In the first stage, HGAA incorporates historical gradient information into the iterative process of generating adversarial samples. It considers gradient similarity between iterative steps to stabilize the updating direction, resulting in improved transfer gradient estimation and stronger adversarial samples. In the second stage, a soft sensor domain-adaptive training model is developed to learn common features from adversarial and original samples through domain-adaptive training, thereby avoiding excessive leaning toward either side and enhancing the adversarial robustness of DLSS without robust overfitting. To demonstrate the effectiveness of DAAT, a DLSS model for crystal quality variables in silicon single-crystal growth manufacturing processes is used as a case study. Through DAAT, the DLSS achieves a balance between defense against adversarial samples and prediction accuracy on normal samples to some extent, offering an effective approach for enhancing the adversarial robustness of DLSS.

Semi-supervised soft sensor development based on dynamic dimensionality reduction-assisted large-scale pseudo label optimization and sample-weighted quality-relevant deep learning

Data-driven soft sensors have become popular tools for estimating critical quality variables in the process industry. However, in practical applications, it is very common that the unlabeled data are abundant but the labeled data are scarce, which poses a great challenge for developing high-performance data-based soft sensors. Thus, a dynamic dimensionality reduction-assisted large-scale pseudo label optimization method (DDR-LSPLO) is proposed for achieving sample expansion. This method repeatedly converts the LSPLO issue into a reduced-dimension pseudo label optimization problem with the low-confidence pseudo labels as new decision variables during the evolutionary optimization process. Meanwhile, to tackle the sample imbalance problem resulting from the inclusion of large-scale pseudo-labeled samples, a sample expansion and weighting-based quality-relevant autoencoder (SEWQAE) is developed for semi-supervised soft sensor modeling. The effectiveness and superiority of the proposed DDR-LSPLO and SEWQAE methods are verified through an industrial chlortetracycline (CTC) fermentation process and a simulated Tennessee Eastman (TE) chemical process.

Dynamic characteristics optimization of elastomer for resistance strain force sensor

PurposeThis paper aims to provide a new method to study and improve the dynamic characteristics of the four-column resistance strain force sensor through the elastomer structure design and optimization.Design/methodology/approachBased on the mechanism analysis method, the authors first present a dynamic characteristic model of the four-column resistance strain force sensors’ elastomer. Then, the authors verified and modified the model according to the Solidworks finite element simulation results. Finally, the authors designed and optimized two types of four-column elastomers based on the dynamic characteristic model and verified the improvement of sensor dynamic performance through a hammer knock dynamic experiment.FindingsThe Solidworks finite element simulation and hammer knock dynamic experiment results show that the relative error of the model is less than 10%, which confirms the accuracy of the model. The dynamic performance of the sensors based on the model can be improved by more than 30%, which is a great improvement in sensor dynamic performance.Originality/valueThe authors first present a dynamic characteristic model of the four-column elastomer and optimize the four-column sensors successfully based on the mechanism analysis method. And a new method to study and improve the dynamic characteristics of the resistance is provided.

Modeling of fiber-optic vibration sensor based on attached bragg grating

The article presents a numerical analysis of the possibility of creating a fiber-optic vibration sensor based on an attached Bragg grating. The sensor consists of two inert masses connected to each other by several membranes and an optical fiber with a fiber Bragg grating (FBG). The study is divided into two main stages: modeling the mechanical part of the sensor, assessing cross-influence and the influence of external temperatures, and developing a signal processing algorithm. A mechanical study of the vibration sensor was conducted to determine its resonant frequency and to determine the optimal length of the elastic beam. The optical part of the study focused on selectingoptimal parameters for the configuration of the arrangement of optical fibers. The results showed that the sensor has its own oscillation frequency above 2000 Hz, and there is a weak dependence of the sensor's sensitivity on the frequency of oscillations with a beam length of 1 mm. The research contributes to the development of fiber-optic accelerometers for measuring and analyzing vibration in various application areas.

Soft sensor for viable cell counting by measuring dynamic oxygen uptake rate

Regulatory authorities in biopharmaceutical industry emphasize process design by process understanding but applicable tools that are easy to implement are still missing. Soft sensors are a promising tool for the implementation of the Quality by Design (QbD) approach and Process Analytical Technology (PAT). In particular, the correlation between viable cell counting and oxygen consumption was investigated, but problems remained: Either the process had to be modified for excluding CO2 in pH control, or complex kLa models had to be set up for specific processes. In this work, a non-invasive soft sensor for simplified on-line cell counting based on dynamic oxygen uptake rate was developed with no need of special equipment. The dynamic oxygen uptake rates were determined by automated and periodic interruptions of gas supply in DASGIP® bioreactor systems, realized by a programmed Visual Basic script in the DASware® control software. With off-line cell counting, the two parameters were correlated based on linear regression and led to a robust model with a correlation coefficient of 0.92. Avoidance of oxygen starvation was achieved by gas flow reactivation at a certain minimum dissolved oxygen concentration. The soft sensor model was established in the exponential growth phase of a Chinese Hamster Ovary fed-batch process. Control studies showed no impact on cell growth by the discontinuous gas supply. This soft sensor is the first to be presented that does not require any specialized additional equipment as the methodology relies solely on the direct measurement of oxygen consumed by the cells in the bioreactor.

Simulation Validation of the Analytical Model of Differential Inductive Servo Sensor for Imaging Signal Generation and Measurement Applications

This paper explores the utility of the seesaw model sensor apparently giving a differential output in a differential manner, which validates one of the major findings of this research using simulation techniques. In this article, simulation is carried out of the designed mathematical model and its necessary derivations for frequency, frequency hysteresis, sensing voltage, magnetic field strength, resonant frequency, mass of bar and mass of target object and the current excitation responses also obtained. Furthermore, the linearity issues of the transducer response curve are addressed for an application scenario of using an inductor with 3R resistor in a differential manner. The parameter of interest is displacement of the core bringing about changes in inductance. Displacement makes the differential inductor to oscillate and undergo changes in differential style. Other parameters of interest such range, frequency, inward output voltage, outward output voltage were explored and results obtained validated the analytical modelling from mathematical derivations in recent literature.

The present state of low-cost air quality sensors in Japan and their accuracy

ABSTRACT The Sustainable Development Goals (SDGs) emphasise the importance of clean air (SDG 3 and SDG 11.6). Japan faces challenges in air quality prediction, especially in rural areas, because of the absence of a climate-specific calibration model and limited deployment of low-cost sensors. Our study compared three sensor types: low-cost pre-assembled and pre-calibrated sensors (PurpleAir); self-assembled low-cost sensors (Citizen Science Project); and highly accurate weather stations, focusing on PM2.5 values. Linear and generalised linear mixed models show potential for low-cost sensor calibration.

Strategy to expedite gas sensor calibration based on cyclic temperature operation and data augmentation

The gas sensor calibration using data-driven approach requires a large volume of training data. The conventional method demands numerous measurements of response data, consuming substantial manpower and time resources. To expedite and streamline the sensor calibration and model deployment, this study proposes a cost-effective and practical calibration strategy combining data augmentation and data preprocessing. This strategy aims to minimize both the collection time and amount of measurement data necessary while ensuring high identification accuracy and generalization performance. The random cropping technique based on sliding sampling window is developed to expand the sample size by randomly and consecutively resampling new data from the limited raw response data. This approach allows any fixed-length sequence from the raw response curve to serve as the training and test sample, facilitating fast and real-time measurement for online detection and dynamic analysis. Furthermore, data preprocessing techniques using sequence normalization and fast Fourier transform are employed to extract the species and concentration features, thereby mitigating drift-like effects in the response data. The effectiveness of this strategy is demonstrated with real measurement data, showcasing significant improvements in identification accuracy and generalization performance compared to conventional methods.

On the resonance frequency of the irregular shaped visco-acoustic Helmholtz cavity problem

In this paper we develop a closed-form mathematical solution for the resonance frequency of the axisymmetric irregular shaped visco-acoustic Helmholtz cavity problem. The solution is based on a conformal mapping that transforms the governing visco-acoustic equations of motion for the irregular cavity geometry into a new domain described by a set of orthogonal curvilinear coordinates. These orthogonal curvilinear coordinates have an important characteristic relative to the transformation of the physical boundary: the mapped domain is constructed such that the locus of all points for the length-wise profile of the physical boundary coincides with a constant value for a (quasi radial) coordinate in the new mapped domain. In these orthogonal curvilinear coordinates we obtain an analytical solution for the visco-acoustic equations of motion and exactly satisfy the no-slip fluid boundary conditions over the irregular shaped cavity walls by matching the coefficients of an orthogonal Legendre polynomial series representation for the non-uniform velocity on the cavity walls. The mathematical solution is an approximation to the extent that the scale factors chosen for the mapping consist of strictly the mutual-coupling terms of the true scale factors. Thus we analyze some results to understand the correlation of the analytical solution frequency response spectra with numerical finite element analyses of the idealized boundary conditions with various irregular shaped cavity geometries. The results for fundamental resonance frequency were found to exhibit good correlation with the results of finite element analyses over a significant range of fluid properties with different irregular cavity geometries. As the primary motivation for the derivation stems from the lack of a rapid yet accurate alternative to computationally intensive MultiPhysics FEA in the formulation of acoustic trendlines for sensor development, the solution’s acoustic impedance was applied as dispersive loading in a simple lumped parameter model of a fluid identification sensor for comparison. The resulting acoustic trendlines of the reduced (5 DOF) model were found to exhibit good correlation with the results of the MultiPhysics FEA (>105 DOF). Thus the solution was found to enable a closed-form analytical method for rapid construction of parametric studies in the practical design of fluid analysis sensors based on complex resonant cavity geometries.

Laplacian regularization of linear regression model for semi-supervised industrial soft sensor development

With more and more data collected from modern industrial processes, data-based soft sensor modeling methods have become popular for online prediction of those difficult-to-measure key variables. As one of the most widely used methods, the linear regression model has been studied for many years. In this paper, a new semi-supervised form of data-based soft sensor is developed upon this simple but commonly used linear regression model, in order to make it more engineering applicable. Particularly, a semi-supervised model structure is formulated by introducing a Laplacian regularization term, through which a lot more unlabeled data samples can be well incorporated for modeling. With an improved estimation for the input data distribution, the real data features can be better captured and thus the model parameters are well restricted, in order to provide a better performance. In addition, the basic Laplacian regularized linear regression model is further extended to nonlinear counterparts by the introductions of basis function and kernel trick in the input data space. For performance evaluation, two industrial case studies are provided, based on which both feasibility and effectiveness of the developed method are confirmed.

Using high sensitivity DC accelerometers for torque estimation on a wind turbine gearbox

In this paper we propose a novel approach to estimate the torque on a wind turbine gearbox using single-axis high sensitivity DC accelerometers. The torque estimator uses an Augmented Extended Kalman Filter (AEKF), combining physics-based models with measurement data. The accelerometer sensor model is derived by analysing the measurements, leading to the identification of two torque-driven acceleration contributions: acceleration due to force excitation and acceleration due to a change in measurement direction. For the latter, two different sources, with distinct frequency content, are capable of changing the measurement direction. The first source corresponds to rigid body motion (f < 0.15Hz), the second one to ring gear deformation (0.15 < f < 4Hz). Using this information, the measured signal can be filtered to target specific acceleration contributions. The first torque estimation approach targets frequencies from 0.15 to 4Hz and produces a Normalised Mean Absolute Error (NMAE) of 2.12%. The second approach targets frequencies up to 0.15Hz, and will estimate the rigid body deflection angle. The torque is calculated through a linear identified relation between the rigid body angle and applied input torque. This approach yields an NMAE of 2.59%. This leads to the conclusion that high sensitivity DC accelerometers are a valid option for estimating torque on a wind turbine gearbox. These torque monitoring approaches will pave the way for advanced condition monitoring strategies and the calculation of remaining useful lifetime of the gearbox.

Research on S-type fiber Bragg grating acceleration sensor for low-frequency vibration monitoring.

The acquisition of low-frequency vibration signals in earthquake exploration is a crucial aspect of seismic exploration technology. Addressing the difficulty current fiber optic sensors face in monitoring low-frequency vibrations, this paper proposes an acceleration sensor based on S-type Fiber Bragg Grating (FBG). First, an FBG acceleration sensor model is established and theoretically analyzed. Then, the impact of structural parameters on the sensor's sensitivity and natural frequency is examined, employing the ANSYS finite element analysis software for static stress and modal simulation analysis. Finally, a prototype is developed, and a low-frequency vibration testing system is assembled to test the sensor's performance. The results show that the sensor has a natural frequency of 34Hz, operates in a frequency band of 0.2-14Hz, has a dynamic range of 63.5 dB, a lateral interference of less than 3%, a sensitivity of ∼274.45 pm/g, good linearity, and is insensitive to temperature. The findings provide a reference for the development of similar sensors and further exploration of the lower frequency limit.

Correlation of digital twin and roll surface sensor results for AZ31 alloy TRC process

Due to the growing interest in lightweight constructions, the continuous casting of nonferrous metals is continuously developing as a result of the cost-effectiveness of this process, which combines several stages of sheet production. Unfortunately, because of the characteristics of the process, the parameters in the roll gap, such as, for example, pressure and temperature, are unknown, significantly affects the understanding of the phenomena occurring in the material during rolling. Therefore, at IMF Freiberg, a sensor consisting of a piezo sensor and two thermocouples measuring the temperature at two different heights was mounted on the surface of the roll, making it possible to control the process parameters live during the TRC trial. The measurements were further supported by a digital twin in the form of a layer model, combining a viscous and solid region for each layer in a single tool. The computations in this tool are performed offline and the computation time is in the order of seconds, thus much less than that of the finite element method. Because the layer model measures the temperature of the magnesium strip, FEM simulations were used to validate measurements from thermocouples. Experimental results have been obtained that allow for a direct correlation between the development of the pressure and temperature and the length of the fully solidified LD part in the roll gap zone, which correlates directly with the effective total equivalent strain. Using the sensor and layer model, it is possible to train a digital twin that can be used for online estimation of the final strip properties obtained in the TRC process.Graphical abstract

A Soft Sensor for Flow Estimation and Uncertainty Analysis Based on Artificial Intelligence: A Case Study of Water Supply Systems

The fourth industrial revolution has transformed the industry, with information technology playing a crucial role in this shift. The increasing digitization of industrial systems demands efficient sensing and control methods, giving rise to soft sensors that have the potential to replace traditional physical sensors in order to reduce costs and enhance efficiency. This study explores the implementation of an artificial neural network (ANN) based soft sensor model in a water supply system to predict flow rates within the system. The soft sensor is centered on a Long Short-Term Memory (LSTM) artificial neural network model using Monte Carlo dropout to reduce uncertainty and improve estimation performance. Based on the results of this work, it is concluded that the proposed soft sensor (with Monte Carlo dropout) can predict flow rates more precisely, contributing to the reduction in water losses, as well as cost savings. This approach offers a valuable solution for minimizing water losses and ensuring the efficient use of this vital resource. Regarding the use of soft sensors based on LSTM neural networks with a careful choice of Monte Carlo dropout parameters, when compared to the multilayer perceptron model, the LSTM model with Monte Carlo dropout showed better mean absolute error, root mean square error, and coefficient of determination: 0.2450, 0.3121, and 0.996437 versus 0.2556, 0.3522, and 0.9954. Furthermore, this choice of Monte Carlo dropout parameters allowed us to achieve an LSTM network model capable of reducing uncertainty to 1.8290, keeping the error metrics also at low levels.

Sensorial Investigation for Automated Patient Data Collection in Hospital Triage: An Exploratory Assessment of the Literature

This paper delves into the proposal of integrating a comprehensive system of vital signs sensors to advance the BayIEEEmax project, which aims to develop an autonomous robot to aid in the triage process within hospital settings. A thorough investigation was conducted to explore commercially available sensor types capable of measuring patients' body temperature, heart rate, and blood pressure. The research focused on assessing the identified sensor models' feasibility and compatibility with the prototype's requisite automation criteria. Ultimately, the selected set of sensors will be incorporated into a printed circuit board and integrated into the physical model of the robot for experimental validation.