Maths Halstensen

New developments in acoustic chemometric prediction of particle size distribution-`the problem is the solution'

We present a new prototype acoustic chemometric approach for prediction of powder particle size distributions, intended for in‐line implementation. The standard basic solutions demand that calibration be carried out on representative, ‘non‐segregated’ reference powder samples. However, as practical powder flow with no segregation is extremely difficult to achieve with the precision needed for calibration, there will always be a significant uncertainty in the reference values relative to what is actually measured. The problem is flow segregation. In order to solve this problem, we have designed a completely new acoustic chemometric approach, which by way of contrast forces the flowing powder mass to segregate as much as possible by various mechanical means. The new approach measures the acoustic signals from an integrated series of segregated, part‐sample characteristics. The calibration X‐data matrix now becomes a three‐way matrix, which demands a three‐way calibration solution to ‘unscramble’ the latent information in the maximally segregated powder sample. Thus the problem is now the solution. Our earlier forays into this matter, which were based on two‐way calibrations, have all been limited by a severe ‘particle size ratio’ bracket outside which destructive self‐damping has effectively negated practical, useful accuracy and precision. The new approach allows a much greater range of contrasting particle sizes. Our first‐generation results achieved by using three‐way PLS‐R as well as the standard two‐way calibrations show that it is more precise than all earlier attempts and can be used for many‐component mixtures without extensive further modifications. We also look at the feasibility of quantifying for prediction of in‐line particle size distributions in an industrial environment. Copyright

Acoustic chemometrics for fluid flow quantifications—II: a small constriction will go a long way

A new approach for non‐invasive quantitative measurement of volume flow rate, multicomponent mixture concentrations as well as density and other physico‐chemical intensive parameters of liquid mixtures flowing in pipelines is presented, based on novel application extensions of the well‐known orifice plate principle (extensively used for flow measurement in pipes). By deliberately transgressing the conventional usage limits, the orifice plate configuration may now also be used for a range of new measurement types, all based on acoustic sensor technology. R&D has been carried out since 1987 by Sensorteknikk A/S and since 1994 in collaboration with ACRG.

Estimation of volume fraction and flow regime identification in inclined pipes based on gamma measurements and multivariate calibration

A combination of gamma measurements and multivariate calibration was applied to estimate multiphase flow mixture density and to identify flow regime. The experiments were conducted using recombined hydrocarbon fluids sampled from an onshore receiving terminal including hydrate thermodynamic inhibitors (monoethylene glycol and methanol (MeOH)). These hydrate inhibitors were added to deionised water at 60% concentration by volume. The experiments were conducted at a temperature of 0 °C and a 75‐bar pressure, comparable with deep water production on the Norwegian continental shelf. Two angles of inclination (1° and 5°) and two water cuts (15% and 85%) were investigated. A single‐energy gamma densitometer was installed on the test facility for measuring the mixture density, whereas the dual‐energy gamma densitometer was traversed linearly from the bottom to the top of the pipe for multivariate calibration and prediction. Seventy partial least square prediction models were calibrated based on single‐phase experimental data. These models were used in estimating the mixture density and identifying the flow regime in all the experiments. The estimated mixture densities were accurate as compared with those from the single‐energy gamma densitometer with the root mean square error of prediction of 13.6 and 9.7 kg/m3 for 1° angle of inclination and 17 and 26.6 kg/m3 for 5° pipe inclination. The models were also able to identify the flow regimes investigated for both 1° and 5° angles of inclination. Copyright

Dynamic system multivariate calibration with low-sampling-rate y data

When the data in principal component regression (PCR) or partial least squares regression (PLSR) form time series, it may be possible to improve the prediction/estimation results by utilizing the correlation between neighboring observations. The estimators may then be identified from experimental data using system identification methods. This is possible also in cases where the response variables in the experimental data are sampled at a low and possibly irregular rate, while the regressor variables are sampled at a higher rate. After a discussion of the options available, the paper shows how the autocorrelation of the regressor variables in such multirate sampling cases may be utilized by identification of parsimonious output error (OE) estimators. An example using acoustic power spectrum regressor data is finally presented. Copyright

On-line near infrared monitoring of ammonium and dry matter in bioslurry for robust biogas production: a full-scale feasibility study.

Heterogeneous substrates fed into agricultural biogas plants originate from many sources with resulting quality fluctuations potentially inhibiting the process. Biogas yield can be substantially increased by optimisation of the organic dry matter load. In this study, near infrared (NIR) spectroscopy was applied on-line in a re-circulating loop configuration operating identically as a full-scale setup. Ammonium could be modelled in the industrially-relevant range 2.42–8.52 gL−1 with an excellent accuracy and precision, slope ∼1.0, r2 = 0.97, corresponding to a relative root mean square error of prediction (RMSEP) of 6.7%. Also, dry matter in the similar plant relevant range 5.8–10.8 weight-percent could be predicted with acceptable accuracy (slope ∼1.0, r2 = 0.83, and a relative RMSEP below 8.0%. Based on these performance characteristics, it was concluded that NIR spectroscopy can be applied for optimising the efficiency of current and future biogas plants, as well as in biorefinery operations converting heterogeneous bioslurry, energy crops, and wastes into value-added products.

Model Choice and Squared Prediction Errors in PLS Regression

Squared prediction errors (SPE) in

A feasibility study of real-time monitoring techniques for scale deposition thickness in pneumatic conveying pipelines

{\bf X}

Equilibrium Measurements of the NH3-CO2-H2O System: Speciation Based on Raman Spectroscopy and Multivariate Modeling

are discussed in relation to the conventional PLSR versus bidiagonalization model and algorithm issue concerning residual and prediction consistency, with focus on process monitoring and fault detection. Our analysis leads to the conclusion that conventional PLSR based on the NIPALS algorithm is ambiguous in SPE values caused by process faults. The basic reason for this is that the sample residuals are not found as projections onto the orthogonal complement of the space where the scores and regression solution are located, and where also the statistical

Multi sensor suite comprising active and passive ultrasonic transducers for monitoring solids flow and silo integrity

{\it T}^{\rm 2}

Acoustic chemometric at-line characterization for monitoring particle size fractions in pneumatically conveyed biomass

limit is defined. The alternative non‐orthogonalized PLSR and bidiagonalization (Bidiag2) algorithms, as well as a simple re‐formulation of the NIPALS algorithm (RE‐PLSR), give unambiguous SPE values, and the last two of these also retain orthogonal score vectors. While prediction results from all of these methods in theory are identical, our conclusion is that methods where the