High-Rate Sulfate Removal Coupled to Elemental Sulfur Production in Mining Process Waters Based on Membrane-Biofilm Technology.

It is anticipated that copper mining output will significantly increase over the next 20 years because of the more intensive use of copper in electricity-related technologies such as for transport and clean power generation, leading to a significant increase in the impacts on water resources if stricter regulations and as a result cleaner mining and processing technologies are not implemented. A key concern of discarded copper production process water is sulfate. In this study we aim to transform sulfate into sulfur in real mining process water. For that, we operate a sequential 2-step membrane biofilm reactor (MBfR) system. We coupled a hydrogenotrophic MBfR (H2-MBfR) for sulfate reduction to an oxidizing MBfR (O2-MBfR) for oxidation of sulfide to elemental sulfur. A key process improvement of the H2-MBfR was online pH control, which led to stable high-rate sulfate removal not limited by biomass accumulation and with H2 supply that was on demand. The H2-MBfR easily adapted to increasing sulfate loads, but the O2-MBfR was difficult to adjust to the varying H2-MBfR outputs, requiring better coupling control. The H2-MBfR achieved high average volumetric sulfate reduction performances of 1.7-3.74 g S/m3-d at 92-97% efficiencies, comparable to current high-rate technologies, but without requiring gas recycling and recompression and by minimizing the H2 off-gassing risk. On the other hand, the O2-MBfR reached average volumetric sulfur production rates of 0.7-2.66 g S/m3-d at efficiencies of 48-78%. The O2-MBfR needs further optimization by automatizing the gas feed, evaluating the controlled removal of excess biomass and S0 particles accumulating in the biofilm, and achieving better coupling control between both reactors. Finally, an economic/sustainability evaluation shows that MBfR technology can benefit from the green production of H2 and O2 at operating costs which compare favorably with membrane filtration, without generating residual streams, and with the recovery of valuable elemental sulfur.

Tuning Generalized Predictive PI controllers for process control applications.

Predictive PI (PPI) controllers have demonstrated to exceed traditional PID controllers when they are applied to systems with long delays. This work proposes a new controller structure and tuning that we call Generalized Predictive PI (GPPI) controller which provides greater design flexibility than PI and PPI strategies. To realize a fair comparison, the design and tuning rules for discrete PI and PPI controllers were developed using optimal arguments based on the root-locus, for critically damped response before a step change in the reference. Experimental results, using industrial equipment, have illustrated the tuning methodology and the performance of the proposed controller under real conditions. Flow and water level process in a laboratory flume were considered. For these processes, First Order Plus Time Delay (FOPTD) models are used. The GPPI control results are encouraging, reducing the settling time plus a very small overshoot before step change in the reference regarding the PI and PPI strategies, up to 41.03% for the flow control loop and up to 54.21% for the level control loop. The discrete analysis of the strategies in the Z plane was performed, allowing for a direct translation to recursive equations that can then be programmed into a Programmable Logic Controller (PLC), other industrial controllers such as Distributed Control Systems (DSC), or microcontrollers, such as Arduino, Raspberry or FPGA. This is an important result, since it demonstrates that the increased complexity of the proposed controller does not hamper its implementation in industrial controller systems. In this work, we used a Rockwell ControlLogix \protect \relax \special {t4ht=®} PLC with Structured Text programming language.

Laser induced breakdown spectroscopy for monitoring the molten phase desulfurization process of blister copper.

The development of real-time monitoring sensors for pyro-metallurgical processes is an analytical challenge, mainly due to adverse environmental conditions, high spectral interferences and multiphase (molten and gas) reactions. This work demonstrates the suitability of stand-off LIBS (ST-LIBS) for real time monitoring of the desulfurization of blister copper which is carried out in molten phase. Here sulfur is removed by the formation of SO2 by supplying oxygen in molten phase. Using ST-LIBS the relative emission intensities of Cu(I) at 351.06 nm, O at 777.34 nm and S at 921.29 nm in both molten and gaseous phase were considered simultaneously during the process. This was possible only by the use high energy laser pulse over up to 270 mJ per pulse. In the case of copper, the selection of emission lines was assessed considering non-linear behavior, which is caused by self-absorption. For the first time, real time determination of sulfur in ppm range is reported by ST-LIBS using low sensitive lines from the NIR region. These results were validated with differential optical absorption spectroscopy (DOAS) as gold standard method. The analytical information obtained by LIBS can precisely determine the critical end-point of the desulfurization where the removal of sulfur is finished, and copper started to oxidize.

Combining prior knowledge with input selection algorithms for quantitative analysis using neural networks in laser induced breakdown spectroscopy.

Laser-induced breakdown spectroscopy (LIBS) is an emerging technique for the analysis of rocks and mineral samples. Artificial neural networks (ANNs) have been used to estimate the concentration of minerals in samples from LIBS spectra. These spectra are very high dimensional data, and it is known that only specific wavelengths have information on the atomic and molecular features of the sample under investigation. This work presents a systematic methodology based on the Akaike information criterion (AIC) for selecting the wavelengths of LIBS spectra as well as the ANN model complexity, by combining prior knowledge and variable selection algorithms. Several variable selection algorithms are compared within the proposed methodology, namely KBest, a least absolute shrinkage and selection operator (LASSO) regularization, principal component analysis (PCA), and competitive adaptive reweighted sampling (CARS). As an illustrative example, the estimation of copper, iron and arsenic concentrations in pelletized mineral samples is performed. A dataset of LIBS emission spectra with 12 287 wavelengths in the range of 185-1049 nm obtained from 131 samples of copper concentrates is used for regression analysis. An ANN is then trained considering the selected reduced wavelength data. The results are satisfactory using LASSO and CARS algorithms along with prior knowledge, showing that the proposed methodology is very effective for selecting wavelengths and model complexity in quantitative analyses based on ANNs and LIBS.

Comparative Analysis of a Principal Component Analysis-Based and an Artificial Neural Network-Based Method for Baseline Removal.

This work presents a non-parametric method based on a principal component analysis (PCA) and a parametric one based on artificial neural networks (ANN) to remove continuous baseline features from spectra. The non-parametric method estimates the baseline based on a set of sampled basis vectors obtained from PCA applied over a previously composed continuous spectra learning matrix. The parametric method, however, uses an ANN to filter out the baseline. Previous studies have demonstrated that this method is one of the most effective for baseline removal. The evaluation of both methods was carried out by using a synthetic database designed for benchmarking baseline removal algorithms, containing 100 synthetic composed spectra at different signal-to-baseline ratio (SBR), signal-to-noise ratio (SNR), and baseline slopes. In addition to deomonstrating the utility of the proposed methods and to compare them in a real application, a spectral data set measured from a flame radiation process was used. Several performance metrics such as correlation coefficient, chi-square value, and goodness-of-fit coefficient were calculated to quantify and compare both algorithms. Results demonstrate that the PCA-based method outperforms the one based on ANN both in terms of performance and simplicity.

Flame spectra-temperature estimation based on a color imaging camera and a spectral reconstruction technique.

In this paper a low-cost, practical pixel-based flame spectrum and temperature estimation system based on flame color images is proposed. A spectral resolution of ∼ 0.4 nm is achieved with an optical system formed by a color camera, a linear model, a flame's spectral training data, and a spectral reconstruction procedure. As a proof of concept, the estimated spectra are compared to local measurements performed with a commercial spectrometer. In order to estimate the absolute flame-temperature maps, two radiometric images at different wavelengths are reconstructed and the two-color pyrometry method is applied. Experiments show errors of about 2.0% over the estimated temperature, making this system a practical tool for flame sensing in combustion-process monitoring.

Removing baseline flame's spectrum by using advanced recovering spectrum techniques.

In this paper, a novel automated algorithm to estimate and remove the continuous baseline from measured flame spectra is proposed. The algorithm estimates the continuous background based on previous information obtained from a learning database of continuous flame spectra. Then, the discontinuous flame emission is calculated by subtracting the estimated continuous baseline from the measured spectrum. The key issue subtending the learning database is that the continuous flame emissions are predominant in the sooty regions, in absence of discontinuous radiation. The proposed algorithm was tested using natural gas and bio-oil flames spectra at different combustion conditions, and the goodness-of-fit coefficient (GFC) quality metric was used to quantify the performance in the estimation process. Additionally, the commonly used first derivative method (FDM) for baseline removing was applied to the same testing spectra in order to compare and to evaluate the proposed technique. The achieved results show that the proposed method is a very attractive tool for designing advanced combustion monitoring strategies of discontinuous emissions.

Analysis and compensation of the effects of analog VLSI arithmetic on the LMS algorithm.

Analog very large scale integration implementations of neural networks can compute using a fraction of the size and power required by their digital counterparts. However, intrinsic limitations of analog hardware, such as device mismatch, charge leakage, and noise, reduce the accuracy of analog arithmetic circuits, degrading the performance of large-scale adaptive systems. In this paper, we present a detailed mathematical analysis that relates different parameters of the hardware limitations to specific effects on the convergence properties of linear perceptrons trained with the least-mean-square (LMS) algorithm. Using this analysis, we derive design guidelines and introduce simple on-chip calibration techniques to improve the accuracy of analog neural networks with a small cost in die area and power dissipation. We validate our analysis by evaluating the performance of a mixed-signal complementary metal-oxide-semiconductor implementation of a 32-input perceptron trained with LMS.

Photodiode-based sensor for flame sensing and combustion-process monitoring.

A nonintrusive low-cost sensor based on silicon photodiode detectors has been designed to analyze the formation and behavior of excited CH(*) and C(2)(*) radicals in the combustion process by sensing the spectral emission of hydrocarbon flames. The sensor was validated by performing two sets of experiments for both nonconfined and confined flames. For a nonconfined oil flame, the sensor responses for the axial intensity were highly correlated with the measurements obtained with a radiometer. For confined gas flames the ratio between the signal corresponding to C(2)(*) and CH(*) was successfully correlated with the CO pollutant emissions and the combustion efficiency. These results give additional insight on how to prevent an incomplete combustion using spectral information. The fast response, the nonintrusive character, and the instantaneous measurement of the needed spectral information makes the proposed optical sensor a key element in the development of advanced control strategies for combustion processes.

Analysis of artificial neural networks for pattern-based adaptive control.

Adaptive pattern-based control strategies adapt their parameters from an analysis of response patterns exhibited by the system. This work presents an analysis of a class of artificial neural network (ANN) pattern-based adaptive control. It provides conditions under which the adaptive algorithm will converge, and it also characterizes the closed-loop stability properties. In addition, a method for monitoring the adaptation is also proposed. Several simulation examples illustrate our findings.

An adaptive pattern based nonlinear PID controller.

This paper presents a nonlinear proportional-integral-derivative (PID) controller, combining a pattern based adaptive algorithm to cope with the problem of tuning the controller, and an associative memory to store the parameters, according to different operating conditions. The simplicity of the algorithm enables its implementation in current programmable logic controller technology. Several real-time experiments, carried out in a pressurized tank, illustrate the performance of the proposed controller.