Search

Wastewater flowrate exhibits valuable information on the conditions in a sewer line. However, existing hardware flowrate sensors are rather expensive, vulnerable to fouling and breakages, and require frequent and laborious maintenance. Therefore, they are typically only mounted in a few key locations of a sewer system, leading to a lack of important information in a major part of the network. Utilizing more cost-effective sensors and a soft sensor approach is advisable for estimating the flowrate at locations where hardware sensors are lacking. Here, the development and testing of a data-driven soft sensor for estimating the wastewater flowrate based on the water level information obtained from a low-cost ultrasonic distance sensor are presented. The research included a long-term functionality testing period of the sensor in a cold region. The soft sensor-based flowrate was applied to estimate inflow and infiltration, indicating the conditions of the sewer line. The harsh conditions inside the sewer manhole caused challenges for the reliability of the distance measurement based on an ultrasonic principle. With the developed model-based soft sensor, it seems possible to accurately estimate the wastewater flowrate. Together with additional information, it might also enable accurate monitoring of inflows and infiltrations.

Barkhausen noise (BN) measurements are commonly used for surface characterisation. However, often there is also a need to verify the sub-surface region because detrimental tensile stresses may be present after different manufacturing steps. Especially in a grinding burn, the surface stress may be compressive, but it changes quickly into tensile stress below the surface. The aim of this study was to find out whether regular surface-sensitive BN measurement is also sensitive to the stresses below the surface caused by grinding burns. More specifically, the aim was to study the relationship between BN features and sub-surface stresses and to identify a model that estimates sub-surface stresses. Real samples were collected from an actual process. The samples were cylindrical samples manufactured from commercial alloyed AISI/SAE L6 steel that was through-hardened prior to grinding. Barkhausen noise measurements were carried out for 42 grinding burn locations followed by X-ray diffraction-based residual stress surface measurements and residual stress depth profiles. Depth information was obtained through step-by-step electrolytic removal of thin layers. The stress profiles were pre-processed through interpolation and averaged stress was calculated as a function of depth below the surface. Correlation analysis was carried out to evaluate the relationships between BN features and stress at different depths and among BN features. The main outcome of the analysis was that BN measurement is dominated by the sub-surface tensile stresses rather than the compressive stress at the surface. It was also noticed that BN features form two groups, corresponding to average Barkhausen activity and magnetising field strength leading to maximum Barkhausen activity. Models for stress at different steps were identified systematically. The performance of the models for sub-surface stresses was reasonable with R2 values of around 0.85 and root mean squared error (RMSE) values of around 95 MPa. Based on the results, it is concluded that BN measurement provides information about sub-surface stresses and that stress can be evaluated through straightforward modelling, allowing fast detection of grinding burns.

The performance of meteorological data-based methods to forecast plant diseases strongly depends on temporal weather information. In this paper, a data analysis procedure is presented for finding the optimal starting time for forecasting net blotch density in spring barley based on meteorological data. For this purpose, changes in the information content of typically measured weather variables were systemically quantified in sliding time windows and with additionally generated mathematical transformations, namely with features. Signal-to-noise statistics were applied in a novel way as a metric for identifying the optimal starting time instance and the most important features to successfully distinguish between two net blotch densities during springtime itself. According to the results, the information content of meteorological data used in classifying between nine years with and four years without net blotch reached its maximum in Finnish weather conditions on the 41st day from the beginning of the growing season. Specifically, utilising weather data at 41–55 days from the beginning of the growing season maximises successful forecasting potential of net blotch density. It also seems that this time instance enables a linear classification task with a selected feature subset, since the averages of the metrics in two data groups differ statistically with a minimum 68% confidence level for nine days in a 14-day time window.

Machine learning techniques are commonly used in the vibration-based condition monitoring of rotating machines. However, few research studies have focused on model training from a practical viewpoint, namely, how to select representative training samples and operating areas for monitoring applications. We focus on these aspects by studying training sets with varying sizes and distributions, including their effects on the models to be identified. The analysis is based on acceleration and shaft speed data available from an azimuth thruster of a catamaran crane vessel. The considered machine learning algorithm was previously introduced in another study suggesting it could detect defects on the thruster driveline components. In this work, practical guidance is provided to facilitate its implementation, and furthermore, an adaptive method for training subset selection is proposed. Results show that the proposed method enabled the identification of usable training subsets in general, while the success of the previous approach was case-dependent. In addition, the use of Kolmogorov–Smirnov or Anderson–Darling tests for normal distribution, as a part of the method, enabled selections that covered the operating area broadly, while other tests were unfavorable in this regard. Overall, the study demonstrates that reconfigurable and automated model implementations could be achievable with minor effort.

Digital twins have gained a lot of attention in modern day industry, but practical challenges arise from the requirement of continuous and real-time data integration. The actual physical systems are also exposed to disturbances unknown to the real-time simulation. Therefore, adaptation is required to ensure reliable performance and to improve the usability of digital twins in monitoring and diagnostics. This study proposes a general approach to the real-time adaptation of digital twins based on a mechanism guided by evolutionary optimization. The mechanism evaluates the deviation between the measured state of the real system and the estimated state provided by the model under adaptation. The deviation is minimized by adapting the model input based on the differential evolution algorithm. To test the mechanism, the measured data were generated via simulations based on a physical model of the real system. The estimated data were generated by a surrogate model, namely a simplified version of the physical model. A case study is presented where the adaptation mechanism is applied on the digital twin of a marine thruster. Satisfactory accuracy was achieved in the optimization during continuous adaptation. However, further research is required on the algorithms and hardware to reach the real-time computation requirement.

The effect of moisture content and feasibility of its estimation in granular material was investigated via machine vision. The test scheme consisted of saw dust samples derived from Norway spruce with moisture content adjusted to three distinct levels. The effect of moisture when present as ice or liquid water was compared. The experimental procedure consisted of pouring the saw dust under video camera recording. The equipment setup consisted of a vibrator feeder and custom-built pouring frame. Still images were extracted with fixed sample time from the recording done during the pouring procedure. From the extracted frames the dynamic behavior of cone profile was investigated via statistical means. It was observed that 2nd standardized moment correlated with moisture content, phase of water and their interaction. Furthermore, 4th standardized moment correlated with moisture content and phase. The 3rd moment was inspected qualitatively from which it was observed that wet samples exhibited tendency to build mass accumulation sites with increasing moisture content. Samples where water was present as ice this was observed in a very small scale with all moisture content values. Corroborated by optical microscopy, these correlations were deduced to be due to liquid bridging in the bulk. Moisture content when present as ice was, however, observed to have a drastic effect on the overall cone shape. Based on these findings, a machine vision application could be feasible way to estimate moisture content on-line in thawed saw dust by using statistical parameters in classification decision making. This would enable cost-effective on-line monitoring of moisture content and a control circuit to be designed.

Energy efficiency is increasingly being considered as a critical measure of process performance due to its importance both in production costs and in environmental footprint. In this work, an indirect energy efficiency estimator was developed for the Tennessee Eastman (TE) benchmark process for the first time. The TE model was first modified to provide the reference values of energy efficiency. A sophisticated model selection scheme was then applied to build the surrogate-model. The results indicate reasonable model performance with mean absolute prediction error around 1.7%. The results also highlight the limitations present in the training set, which are, together with other practical implementation issues, discussed in this work.

Supercritical carbon dioxide extraction has been established as a promising and clean technology alternative to conventional separation techniques. Despite a high energy demand of extraction processes, their energy analysis has been scarcely considered. In this study, a supercritical carbon dioxide batch extraction process was modelled through system identification, forming a full simulator of its control loops affecting the energy consumption. The modelling was based on data acquired through systematic approach including experimental design and identification of dynamic process responses and energy consumption. Regression analysis and 12 identified models for subprocesses showed feasible performance during simulations with experimental data. The best local model for a subprocesses exhibited a Mean Absolute Percentage Error of 3% with independent test data. Regression model for steady-state electricity consumption showed a Mean Absolute Percentage Error of 7.6%, also suggesting the existence of nonlinearities between the response and other process variables. The identification approach reveals new information on energy consumption and dynamics of energy consumption of supercritical extraction in transient operating conditions. The models can be applied for further developments in real-time energy monitoring and optimization of supercritical extraction processes.

Catalyst development for biorefining applications involves many challenges. Mathematical modeling can be seen as an essential tool in assisting to explain catalyst performance. This paper presents studies on several machine learning (ML) methods that can model the performance of heterogeneous catalysts with relevant descriptors. A systematic approach for selecting the most appropriate ML method is taken with focus on the variable selection. Regularization algorithms were applied to variable selection. Several different candidate model structures were compared in modeling with interpretation of results. The systematic modeling approach presented aims to highlight the necessary tools and aspects to unexperienced users of ML. Literature datasets for the hydrogenation of 5-ethoxymethylfurfural with simple bimetal catalysts, including main metals and promoters, were studied with the addition of catalyst descriptors found in the literature. Good results were obtained with the best models for estimating conversion, selectivity, and yield with correlations between 0.90 and 0.98. The best identified model structures were support vector regression, Gaussian process regression, and decision tree methods. In general, the use of variable selection procedures was found to improve the performance of models. The modeling methods applied thus seem to exhibit a strong potential in aiding catalyst development based mainly on the information content of descriptor datasets.

Application of membrane technologies in biorefinery processes has been studied for some time. The heterogenous nature of biorefinery steams, however, results in unideal performance of membrane systems and considerable fouling of membranes, which is decreasing the efficiency of separation. As a part of BioSPRINT project, this study focuses on application of separating monomeric sugars from the hemicelluloses fraction of lignocellulosic biomass, where pressure-driven nanofiltration with several diafiltration stages has been proposed for the separation task. Diafiltration is required to overcome the decreased separation efficiency when the retentate concentrations and viscosity increases. A lumped parameter dynamical model of the diafiltration plant is applied. The key model parameters are identified from experimental data from a laboratory membrane unit to reflect the considered biorefinery process. The model is then simulated to study the sensitivity of the uncertain model parameters (related to membrane fouling, solute concentrations, viscosity, and mass transfer coefficients) to the diafiltration plant performance (product purity, operation time). The model is implemented in the MATLAB®/Simulink environment. The simulation results are expected to identify potential sources of scale-up challenges in biorefinery-related membrane applications. The developed dynamic model also allows to investigate different operational strategies of diafiltration plants in the future.

Biomethane is a renewable product that can directly substitute its fossil counterpart, although its synthesis from residual biomasses has some hurdles. Because of the complex nature of both biomasses and the microbial consortia involved, innovative approaches such as mathematical modeling can be deployed to support possible improvements. The goal of this study is two-fold, as we aimed to modify a part of the Anaerobic Digestion Model No. 1 (ADM1), describing biomethane production from activated sludge, matching with its actual microbial nature, and to use the model for identifying relevant parameters to improve biomethane production. Firstly, thermodynamic analysis was performed, highlighting the direct route from glucose to biomethane as the most favorable. Then, by using MATLAB® and Simulink Toolbox, we discovered that the model fails to predict the microbiological behavior of the system. The structure of the ADM1 model was then modified by adding substrate consumption yields in equations describing microbial growth, to better reflect the consortium behavior. The updated model was tested by modifying several parameters: the coefficient of decomposition was identified to increase biomethane production. Approaching mathematical models from a microbiological point of view can lead to further improvement of the models themselves. Furthermore, this work represents additional evidence of the importance of informatics tools, such as bioprocess simulations to foster biomethane role in bioeconomy.

In modern agriculture, the pesticides and the need to decrease their use is under discussion. Optimization methods and modelling tools are important research areas in this context. In this paper, data analysis, feature generation and selection in moving windows have been utilized for the evaluation of net blotch risk in barley. Two different datasets: The open data from the Finnish Meteorological Institute and the historical observation of the net blotch severity in different fields in Finland are combined with feature generation techniques. T-test is then applied to select the most statistically suitable features for prediction the net blotch risk from weather measurements. Analysis proceeds in moving data windows to indicate the most informative time period to predict the risk of net blotch during the growing season. Results show that the selection of the proper time instance and the length of data window may enhance strongly the potential performance of prediction methods for risk analysis on plant disease.

Digital twins for performance-oriented applications in industrial environments require systematic model maintenance. Model adaptation requires efficient optimization tools and continuous evaluation of measurement quality. The adaptation and model performance evaluation are based on the modeling error, making the adaptation prone also to the measurement errors. In this paper, a framework for combining model adaptation and measurement quality assurance are discussed. Two examples with simulated industrialscale biopharmaceutical penicillin fermentation are presented to illustrate the usability of the framework.

The energy industry is striving for cost-efficient production with low emissions. The efficiency of energy conversion processes is one of the most important factors affecting the path to such a goal. This paper provides a framework for the monitoring of steam boiler operation in power stations. The framework is based on the use of historical process data as a reference for real-time operation. The actual boiler efficiency is monitored together with its expected efficiency, which is an estimate of the highest historical efficiency in the corresponding process state, based on a data-driven model. In the presented approach, the process state is defined on the basis of variables that have the strongest correlation with boiler efficiency according to information-theoretic variable ranking. Boiler performance is monitored using a statistical process control chart for the difference between the expected and actual efficiencies. The framework was tested using data from a circulating fluidised bed boiler and from a corner-fired boiler. The results revealed that the strongest correlations between process variables and boiler efficiency are substantially consistent in both cases. Moreover, the framework provides a novel measure for boiler performance enhancement.

Implementation of new energy efficiency measures for the heating and building sectors is of utmost importance. Demand side management offers means to involve individual buildings in the optimization of the heat demand at city level to improve energy efficiency. In this work, two models were applied to forecast the heat demand from individual buildings up to a city-wide area. District heating data at the city level from more than 4000 different buildings was utilized in the validation of the forecast models. Forecast simulations with the applied models and measured data showed that, during the heating season, the relative error of the city level heat demand forecast for 48 h was 4% on average. In individual buildings, the accuracy of the models varied based on the building type and heat demand pattern. The forecasting accuracy, the limited amount of measurement information and the short time required for model calibration enable the models to be applied to the whole building stock. This should enable demand side management and lead to the predictive optimization of heat demand at city level, leading to increased energy efficiency.

A novel dynamic model for the temperature inside buildings is presented, aiming to improve energy efficiency by providing predictive information on the heat demand. To analyse the performance and generalizability of the modelling approach, real measurement data was gathered from five different types of buildings. Easily available data from various sources was utilized. The chosen model structure leads to a minimal number of input variables and free parameters. Simulations with real data from five buildings, and applying the identical model structure showed that the average modelling error during the 28-h prediction horizon was constantly below 5%. The results thus demonstrate that the model structure can be standardized and easily applied to predict the indoor temperatures of large buildings. This would finally enable demand side management and the predictive optimization of the heat demand at city level.

An advanced model-based control method for the integrated fuel processing and a fuel cell system consisting of ethanol reforming, hydrogen purification, and a proton exchange membrane fuel cell is presented. For process identification, a physical model of the process chain was constructed. Subsequently, the simulated process was approximated with data-driven control models. Based on these control models, a hierarchical control framework consisting of model predictive controller and a global optimization algorithm was introduced. The performance of the new control method was evaluated with simulations. Results indicate that the new optimization concept enables resource efficient and fast control of the studied energy conversion process. Fast and efficient fuel cell process could then provide sustainable power source for autonomous and mobile applications in the future.

As a part of the H2020 SPIRE project, Intensified by Design (IbD®), mineral beneficiation process chain will be intensified in terms of more efficient usage of energy and raw materials. The work focuses on optimizing control of the grinding circuit, coarse flotation process, new solids measurements and soft-sensors, as well as adaptive dynamic mass balancing. All demonstrators are interconnected through plant-wide optimization. In this paper, the process intensification activities carried out at the mini-pilot scale mineral beneficiation plant are briefly described. The real-time adaptation strategy utilizing Differential Evolution for parameter estimation in a sliding window is presented. The initial results show a successful parameter estimation result even in a short time window.

Easily adaptable indoor temperature and heat demand models were applied in the predictive optimization of the heat demand at the city level to improve energy efficiency in heating. Real measured district heating data from 201 large buildings, including apartment buildings, schools and commercial, public, and office buildings, was utilized. Indoor temperature and heat demand of all 201 individual buildings were modelled and the models were applied in the optimization utilizing two different optimization strategies. Results demonstrate that the applied modelling approach enables the utilization of buildings as short-term heat storages in the optimization of the heat demand leading to significant improvements in energy efficiency both at the city level and in individual buildings.

A nonlinear empirical model between stress and Barkhausen noise is identified in this study. The identification procedure uses a genetic algorithm followed by a Nelder-Mead optimization procedure. The model is identified with the data set where an external load is applied to RAEX400 low alloyed hot-rolled steel samples. The results of the study show that the identified model performs well in stress predictions. The identified model includes three terms which are in accordance with the literature.

A model-based control strategy for the compensation of combustion power fluctuations in small-scale biomass fired boilers is presented. Delay-free heat output model is obtained by utilising combustion temperature measurements. Local principal component regression models are constructed to cope with changing combustion conditions and redundant input variables found by data analysis. The modelling approach was validated using data from 30 kW and 300 kW boilers. Usability of the discussed monitoring method for compensating fuel power disturbances by the primary air control was tested in practice. The results imply that modelling information can be utilised as a part of the control structure, enabling the cost-effective way to reduce emissions and maintain optimal combustion conditions in the distributed energy production with solid biomass fuels.

A novel method for real-time moisture level monitoring of the solid biomass fuel in grate combustion is presented. The measurement principle is based on temperature sensor information from both flame and a fuel bed. Based on the combustion theory and data analysis, selected features have been extracted from the fused sensor information and estimate of the fuel quality is then made continuously with calculated features. The monitoring approach has been tested in a 300 kW stoker combustion unit intended for decentralized heat production, over a wide range of different process conditions and wood fuel moistures, giving satisfactory accuracy for control purposes. The availability of moisture information made possible to adjust primary/secondary air ratio, leading to reduction of emissions and excess air. Based on the results, the method has capability to give new possibilities for cost effective control and more energy efficient use of solid biomass as a fuel in small-scale energy production.

A procedure for surveying process data sets is presented. For this, linear models constructed in varying length, sliding data windows to determine the usefulness of data segments for process identification are utilised. The discussed approach has been applied to an industrial wood debarking plant and a biomass boiler analysis, enabling the preliminary study of process variables and conditions affecting the non-optimal process conditions. In addition, main process interactions and delays were easily discovered from the structures of the interpretable linear model candidates. It is concluded that the analysis can provide valuable information also for modelling and control of continuous processes.

A soft computing monitoring approach for automated screw insertions is presented. A model based monitoring method is developed with systematically collected experimental data and fundamental process knowledge to verify the quality of assemblies. The model for quality monitoring is based on Linguistic Equations (LE)-a non-linear scaling framework for model variables. Fuzzy reasoning and basic statistical methods are combined to interpret the model residuals and faults. Preliminary tests indicate that the proposed method could successfully cope with changes in manufacturing parameters. Based on the results, the method seems to provide valuable information for quality control of the screw insertion task.

Networked control systems (NCS) could be utilised in several industrial applications. However, the variable time delays introduced by the network impair the NCS performance, resulting even in the instability of the controlled process. To mitigate the delay problems, the advantage is taken from model-based, adaptive controllers. This calls for an efficient approach for on-line analysis of measurements applied to update the controller state in NCS. The paper introduces a new adaptive Model Predictive Controller (MPC) capable of compensating for variations in measurement and actuating delays. Weighting factors for delayed measurements and actuators are adjusted based on normalised version of mutual information that is calculated using a procedure described in the paper. The method is superior compared with other, more usual, metrics.

Statistical process control (SPC) is a well-known method to perform quality control. However, this method is not adapted to small batches or prototype production because it usually doesn't provide feedback within a proper timeline. This paper describes a case study in which statistical based methods derived from classical SPC have been modified and implemented as part of a decision support system for a semiconductor process. Quality indexes have been defined to monitor the production, perform fault diagnosis and provide control information to operators. The system has been implemented in an electronics factory producing printed circuit board assemblies (PCBA) for telecommunication systems. With products in their prototype stage, over 40% of the production was defected requiring rework, only few data from non-consecutive measurements was available. Implementing a decision support system in such an environment was therefore a challenging task. Nevertheless, the method described in this paper provided adequate information to adjust the process when needed or to tune the system in case of false alarms. Moreover, line worker being supposed to have an active role in terms of quality; it was important to involve them in the design of the system. A greater interest from their side also improved their own understanding of the process and its problems.

Barkhausen noise (BN) measurements are commonly used for surface characterisation. However, often there is also a need to verify the sub-surface region because detrimental tensile stresses may be present after different manufacturing steps. Especially in a grinding burn, the surface stress may be compressive, but it changes quickly into tensile stress below the surface. The aim of this study was to find out whether regular surface-sensitive BN measurement is also sensitive to the stresses below the surface caused by grinding burns. More specifically, the aim was to study the relationship between BN features and sub-surface stresses and to identify a model that estimates sub-surface stresses. Real samples were collected from an actual process. The samples were cylindrical samples manufactured from commercial alloyed AISI/SAE L6 steel that was through-hardened prior to grinding. Barkhausen noise measurements were carried out for 42 grinding burn locations followed by X-ray diffraction-based residual stress surface measurements and residual stress depth profiles. Depth information was obtained through step-by-step electrolytic removal of thin layers. The stress profiles were pre-processed through interpolation and averaged stress was calculated as a function of depth below the surface. Correlation analysis was carried out to evaluate the relationships between BN features and stress at different depths and among BN features. The main outcome of the analysis was that BN measurement is dominated by the sub-surface tensile stresses rather than the compressive stress at the surface. It was also noticed that BN features form two groups, corresponding to average Barkhausen activity and magnetising field strength leading to maximum Barkhausen activity. Models for stress at different steps were identified systematically. The performance of the models for sub-surface stresses was reasonable with R2 values of around 0.85 and root mean squared error (RMSE) values of around 95 MPa. Based on the results, it is concluded that BN measurement provides information about sub-surface stresses and that stress can be evaluated through straightforward modelling, allowing fast detection of grinding burns. publishedVersion