Application of convolutional neural networks for prediction of disinfection by-products

Fluorescence spectroscopy can provide high-level chemical characterization and quantification that is suitable for use in online process monitoring and control. However, the high-dimensionality of excitation–emission matrices and superposition of underlying signals is a major challenge to implementation. Herein the use of Convolutional Neural Networks (CNNs) is investigated to interpret fluorescence spectra and predict the formation of disinfection by-products during drinking water treatment. Using deep CNNs, mean absolute prediction error on a test set of data for total trihalomethanes, total haloacetic acids, and the major individual species were all < 6 µg/L and represent a significant difference improved by 39–62% compared to multi-layer perceptron type networks. Heat maps that identify spectral areas of importance for prediction showed unique humic-like and protein-like regions for individual disinfection by-product species that can be used to validate models and provide insight into precursor characteristics. The use of fluorescence spectroscopy coupled with deep CNNs shows promise to be used for rapid estimation of DBP formation potentials without the need for extensive data pre-processing or dimensionality reduction. Knowledge of DBP formation potentials in near real-time can enable tighter treatment controls and management efforts to minimize the exposure of the public to DBPs.

Correlation Between Insulation Resistance and Temperature Measurement Error in Type K and Type N Mineral Insulated, Metal Sheathed Thermocouples

Mineral insulated, metal sheathed (MI) Type K and Type N thermocouples are widely used in industry for process monitoring and control. One factor that limits their accuracy is the dramatic decrease in the insulation resistance at temperatures above about 600 °C which results in temperature measurement errors due to electrical shunting. In this work the insulation resistance of a cohort of representative MI thermocouples was characterised at temperatures up to 1160 °C, with simultaneous measurements of the error in indicated temperature by in situ comparison with a reference Type R thermocouple. Intriguingly, there appears to be a systematic relationship between the insulation resistance and the error in the indicated temperature. At a given temperature, as the insulation resistance decreases, there is a corresponding increasingly negative error in the temperature measurement. Although the measurements have a relatively large uncertainty (up to about 1 °C in temperature error and up to about 10 % in insulation resistance measurement), the trend is apparent at all temperatures above 600 °C, which suggests that it is real. Furthermore, the correlation disappears at temperatures below about 600 °C, which is consistent with the well-established diminution of insulation resistance breakdown effects below that temperature. This raises the intriguing possibility of using the as-new MI thermocouple calibration as an indicator of insulation resistance breakdown: large deviations of the electromotive force (emf) in the negative direction could indicate a correspondingly low insulation resistance.

Multispectral Active Infrared Thermo-Reflectometry for Absolute Temperature Measurement by Contactless IR Thermography

Knowledge of material emissivity maps and their true temperatures is of great interest for contactless process monitoring and control with infrared cameras when strong heat transfer and temperature change are involved.In this work, we describe the development of a contactless infrared and multispectral imaging technique based on the pyro-reflectometry approach and a specular model of the material reflection.This approach enables in situ and real-time identification of emissivity fields and autocalibration of the radiative intensity leaving the sample by using a black body equivalent ratio.This is done to obtain the absolute temperature field of any specular material using the infrared wavelength.The proposed method is evaluated at room temperature with several heterogeneous samples covering a large range of emissivity values. From these emissivity fields, raw and heterogeneous measured radiative fluxes are transformed into complete absolute temperature fields.

Combining Simulations and Machine Learning for Efficient Prediction of Process Parameters Evolution in Injection Moulding

In recent years, the emerging technologies in the context of Industry 4.0 have led to novel approaches in process monitoring and control, such as the introduction of Reinforcement Learning and Digital Twins. Consequently, large amounts of data, precise modelling and exhaustive simulations are required. The aim of this work is to propose a methodology based on the technique of backward selection to reduce the number of reference points in the simulation stage of manufacturing processes, enhancing the efficiency of data generation and the simplicity of the simulations. The methodology is proved in the particular case of plastic injection moulding simulations.

Bacteriophages in Biological Wastewater Treatment Systems: Occurrence, Characterization, and Function

Phage bacteria interactions can affect structure, dynamics, and function of microbial communities. In the context of biological wastewater treatment (BWT), the presence of phages can alter the efficiency of the treatment process and influence the quality of the treated effluent. The active role of phages in BWT has been demonstrated, but many questions remain unanswered regarding the diversity of phages in these engineered environments, the dynamics of infection, the determination of bacterial hosts, and the impact of their activity in full-scale processes. A deeper understanding of the phage ecology in BWT can lead the improvement of process monitoring and control, promote higher influent quality, and potentiate the use of phages as biocontrol agents. In this review, we highlight suitable methods for studying phages in wastewater adapted from other research fields, provide a critical overview on the current state of knowledge on the effect of phages on structure and function of BWT bacterial communities, and highlight gaps, opportunities, and priority questions to be addressed in future research.

Application of Convolutional Neural Networks for Prediction of 1 Disinfection By-Products

Fluorescence spectroscopy can provide high-level chemical characterization and quantification that is suitable for use in online process monitoring and control. However, the high-dimensionality of excitation-emission matrices and superposition of underlying signals is a major challenge to implementation. Herein the use of Convolutional Neural Networks (CNNs) is investigated to interpret fluorescence spectra and predict the formation of disinfection by-products during drinking water treatment. Using deep CNNs, mean absolute prediction error on a test set of data for total trihalomethanes, total haloacetic acids, and the major individual species were all < 6 µg/L and represent a significant difference improved by 39% - 62% compared to dense neural networks. Heat maps that identify spectral areas of importance for prediction showed unique humic-like and protein-like regions for individual disinfection by-product species that can be used to validate models and provide insight into precursor characteristics. The use of fluorescence spectroscopy coupled with deep CNNs shows promise to be used for rapid estimation of DBP formation potentials without the need for extensive data pre-processing or dimensionality reduction. Knowledge of DBP formation potentials in near real-time can enable tighter treatment controls and management efforts to minimize the exposure of the public to DBPs.

In-Line Observation of Laser Cladding Processes via Atomic Emission Spectroscopy

Direct metal deposition (DMD) can be used for the cladding of surfaces as well as repairing and additive manufacturing of parts and features. Process monitoring and control methods ensure a consistent quality during manufacturing. Monitoring by optical emission spectroscopy of the process radiation can provide information on process conditions and the deposition layer. The object of this work is to measure optical emissions from the process using a spectrometer and identify element lines within the spectra. Single spectra have been recorded from the process. Single tracks of Co-based powder (MetcoClad21) were clad on an S235 base material. The influence of varying process parameters on the incidence and intensity of element lines has been investigated. Moreover, the interactions between the laser beam, powder jet, and substrate with regard to spectral emissions have been examined individually. The results showed that element lines do not occur regularly. Therefore, single spectra are sorted into spectra including element lines (type A) and those not including element lines (type B). Furthermore, only non-ionised elements could be detected, with chromium appearing frequently. It was shown that increasing the laser power increases the incidence of type A spectra and the intensity of specific Cr I lines. Moreover, element lines only occurred frequently during the interaction of the laser beam with the melt pool of the deposition layer.

From Black Box to Machine Learning: A Journey through Membrane Process Modelling

Membrane processes are complex systems, often comprising several physicochemical phenomena, as well as biological reactions, depending on the systems studied. Therefore, process modelling is a requirement to simulate (and predict) process and membrane performance, to infer about optimal process conditions, to assess fouling development, and ultimately, for process monitoring and control. Despite the actual dissemination of terms such as Machine Learning, the use of such computational tools to model membrane processes was regarded by many in the past as not useful from a scientific point-of-view, not contributing to the understanding of the phenomena involved. Despite the controversy, in the last 25 years, data driven, non-mechanistic modelling is being applied to describe different membrane processes and in the development of new modelling and monitoring approaches. Thus, this work aims at providing a personal perspective of the use of non-mechanistic modelling in membrane processes, reviewing the evolution supported in our own experience, gained as research group working in the field of membrane processes. Additionally, some guidelines are provided for the application of advanced mathematical tools to model membrane processes.

A Survey on the Application of WirelessHART for Industrial Process Monitoring and Control

Industrialization has led to a huge demand for a network control system to monitor and control multi-loop processes with high effectiveness. Due to these advancements, new industrial wireless sensor network (IWSN) standards such as ZigBee, WirelessHART, ISA 100.11a wireless, and Wireless network for Industrial Automation-Process Automation (WIA-PA) have begun to emerge based on their wired conventional structure with additional developments. This advancement improved flexibility, scalability, needed fewer cables, reduced the network installation and commissioning time, increased productivity, and reduced maintenance costs compared to wired networks. On the other hand, using IWSNs for process control comes with the critical challenge of handling stochastic network delays, packet drop, and external noises which are capable of degrading the controller performance. Thus, this paper presents a detailed study focusing only on the adoption of WirelessHART in simulations and real-time applications for industrial process monitoring and control with its crucial challenges and design requirements.

A Study on Intelligent Grinding Systems With Industrial Perspective

The digitization thrust on high value manufacturing and services opens-up new opportunities for ensuring; total system uptime, reliability, and efficiency particularly for mission-critical high value assets. The digitization process evolves intelligent manufacturing systems (IMS) which transforms maintenance into predictive reliability for achieving consistent quality throughout manufacturing process. This article unveils the intelligent grinding systems (IGS) for challenging grinding applications. For a more in-depth understanding and analysis of an entire intelligent grinding system, particular aspects within the system were discussed. These include Grinding Models, Process Design Algorithms, Process Monitoring, Process Control, Feature Extraction and Feature Correlation engines. The main focus, especially in the early 2000s, was mainly database development and parameter selection, which then shifted to process monitoring and control as particular technology advances were made. In the various goals that were investigated, it was evident that researchers were aiming for an online real-time system. This notion was driven by the advances in artificial intelligence and improved monitoring sensors, for example, acoustic emission sensors and even other unusual sensors like microphones for more economical and improved data collection and analysis. Although tremendous strides have been made, a substantial amount of work is still required in achieving a fully-fledged real-time intelligent grinding system. The comprehensive findings on IGS system concludes that the real time process update has been improved from few hours to milliseconds.