[Show abstract][Hide abstract] ABSTRACT: This paper presents the development of a passive ultrasonic monitoring system for the detection of acoustic emission (AE) created by chemical particles striking the inner wall of a reactor vessel. The finite element (FE) code PZFlex was used to analyze the complex interactions between chemical particles and the vessel wall. A 4-layer 2D model was developed comprising a liquid load medium and a glass-oil-glass combination corresponding to the jacketed vessel reactor. The model has been experimentally validated with excellent correlation achieved. The excitation function was derived from Hertz's theory and used as the model stimulus corresponding to particles striking the inner glass wall. Analysis of the FE simulations provided the transducer specifications for a passive ultrasonic monitoring system. The system comprises two transducers with complementary characteristics: narrow bandwidth/high sensitivity; wideband/low sensitivity. Importantly, the sensitivity of the resonant transducer provides discrimination of particle concentration. Moreover, the broader bandwidth of the off-resonant device demonstrates potential for in situ estimation of particle size. The performance afforded by this approach has considerable potential for real-time process monitoring in the chemicals and pharmaceutical industries.
Full-text · Article · Jun 2015 · Sensors and Actuators A Physical
[Show abstract][Hide abstract] ABSTRACT: A specially designed thermal vaporiser was used with a process mass spectrometer designed for gas analysis to monitor the esterification of butan-1-ol and acetic anhydride. The reaction was conducted at two scales: in a 150mL flask and a 1L jacketed batch reactor, with liquid delivery flow rates to the vaporiser of 0.1 and 1.0mLmin(-1), respectively. Mass spectrometry measurements were made at selected ion masses, and classical least squares multivariate linear regression was used to produce concentration profiles for the reactants, products and catalyst. The extent of reaction was obtained from the butyl acetate profile and found to be 83% and 76% at 40°C and 20°C, respectively, at the 1L scale. Reactions in the 1L reactor were also monitored by in-line mid-infrared (MIR) spectrometry; off-line gas chromatography (GC) was used as a reference technique when building partial least squares (PLS) multivariate calibration models for prediction of butyl acetate concentrations from the MIR spectra. In validation experiments, good agreement was achieved between the concentration of butyl acetate obtained from in-line MIR spectra and off-line GC. In the initial few minutes of the reaction the profiles for butyl acetate derived from on-line direct liquid sampling mass spectrometry (DLSMS) differed from those of in-line MIR spectrometry owing to the 2min transfer time between the reactor and mass spectrometer. As the reaction proceeded, however, the difference between the concentration profiles became less noticeable. DLSMS had advantages over in-line MIR spectrometry as it was easier to generate concentration profiles for all the components in the reaction. Also, it was possible to detect the presence of a simulated impurity of ethanol (at levels of 2.6 and 9.1% mol/mol) in butan-1-ol, and the resulting production of ethyl acetate, by DLSMS, but not by in-line MIR spectrometry.
No preview · Article · Nov 2014 · Analytica Chimica Acta
[Show abstract][Hide abstract] ABSTRACT: Information about size and shape of particles produced in various
manufacturing processes is very important for process and product development
because design of downstream processes as well as final product properties
strongly depend on these geometrical particle attributes. However, recovery of
particle size and shape information in situ during crystallisation processes
has been a major challenge. The focused beam reflectance measurement (FBRM)
provides the chord length distribution (CLD) of a population of particles in a
suspension flowing close to the sensor window. Recovery of size and shape
information from the CLD requires a model relating particle size and shape to
its CLD as well as solving the corresponding inverse problem.
This paper presents a comprehensive algorithm which produces estimates of
particle size distribution and particle aspect ratio from measured CLD data.
While the algorithm searches for a global best solution to the inverse problem
without requiring further a priori information on the range of particle sizes
present in the population or aspect ratio of particles, suitable regularisation
techniques based on relevant additional information can be implemented as
required to obtain physically reasonable size distributions. We used the
algorithm to analyse CLD data for samples of needle-like crystalline particles
of various lengths using two previously published CLD models for ellipsoids and
for thin cylinders to estimate particle size distribution and shape. We found
that the thin cylinder model yielded significantly better agreement with
experimental data, while estimated particle size distributions and aspect
ratios were in good agreement with those obtained from imaging.
Full-text · Article · Aug 2014 · Chemical Engineering Science
[Show abstract][Hide abstract] ABSTRACT: A thermal vaporiser has been designed for analysis of liquid streams by a process mass spectrometer normally used for gas analysis. Concentrations of benzene, toluene and o-xylene at mg kg−1 levels in ethanol were determined from continuous vaporisation of the liquid. Ions with m/z values of 39, 57, 73, 77, 78, 91, 92 and 106 were selected and the optimal regression model (multiple linear regression with mean-centring) was found using an automated design of experiments approach to calibration model selection. It was discovered that the linearity of the response allowed excellent calibration to be performed using only four standards (at 0 and 110 mg kg−1 for each of the three analytes) and that there were minimal inter-analyte interferences. The detection limit of benzene, toluene and o-xylene was 0.5, 0.8 and 0.5 mg kg−1, respectively. Average differences between the actual and predicted concentrations, expressed as a percentage of the actual concentrations, for 27–82 mg kg−1 of benzene, toluene and o-xylene were 0.5–1.4%, 0.0–0.4% and 0.3–1.6%, respectively, while the average relative standard deviations were 1.3–2.6%, 1.0–2.5% and 1.1–2.3%, respectively. Detection of 3 mg kg−1 changes in the concentration of each of the analytes (at the 36 mg kg−1 level) was also demonstrated, indicating the sensitivity of the technique and the potential ability of the procedure to detect minor deviations in the specification of process streams from continuous analysis.
No preview · Article · Aug 2014 · Analytical methods
[Show abstract][Hide abstract] ABSTRACT: Transmission near-infrared (NIR) measurements of a 1 mm thick aspirin disk were made at different positions as it was moved through a stack of eight 0.5 mm thick disks of microcrystalline cellulose (Avicel). The magnitude of the first derivative of absorbance for the aspirin interlayer at 8934 cm(-1) was lower when the disk was placed at the top or bottom of the stack of Avicel disks, with the largest signal observed when the aspirin was positioned at the central positions. The variation in signal with depth is consistent with that observed previously for transmission Raman spectrometry. In both cases, the trend observed can be attributed to lower photon density at the air-sample interface, relative to the center of the sample, owing to loss of photons to the air. This results in a reduction in the number of photons absorbed or Raman photons generated and subsequently detected when the interlayer occupies a near-surface position.
Full-text · Article · Mar 2014 · Applied Spectroscopy
[Show abstract][Hide abstract] ABSTRACT: The effects of pressure filtration and vacuum agitated drying on cellobiose octaacetate (COA) particles in methanol slurries were studied by making Raman measurements through the glass wall at the side of a filter drier beneath the oil jacket. The change in intensity of methanol peaks in the spectra allowed the removal of the solvent from the particle bed to be monitored. Also, drying curves for COA generated from the Raman measurements gave an indication of the changing physical status of the particle bed during continuous or intermittent agitation. The intensity of the Raman signal for COA depended on the bulk density of the particle bed, which changed due to aggregation and attrition that occurred during solvent removal and particle motion induced by agitation during vacuum drying. Loss on drying (LOD) measurements of samples removed at the end of the pressure filtration and vacuum agitated drying stages established the degree of wetness and confirmed the end point of drying (<0.5% w/w solvent), respectively. Dynamic image analysis confirmed that minimum attrition of COA was achieved when (a) the majority of the methanol was removed during pressure filtration at 0.5bar N2 and (b) intermittent agitation was applied during the vacuum drying stage.
Full-text · Article · Sep 2013 · Chemical Engineering Science
[Show abstract][Hide abstract] ABSTRACT: Biotransformation processes have become industrially important in recent years as routes to the manufacture of high value chemical intermediates. However, measurements of key process features and analyte concentrations during these processes are still typically carried out using off-line analysis methods. Vibrational spectroscopic techniques have been extensively utilised for the monitoring and control of a variety of industrial processes. Despite the techniques success with a range of challenging biological matrices, including fermentation and cell culture systems, application of this approach to biotransformation systems has been limited. In the present study the potential of mid infrared spectroscopy to monitor an industrially relevant de-racemization biotransformation process has been investigated. This process presents a number of difficulties due to the optically challenging sample media, close structural similarities and stoichiometric relationship between the key analytes of interest. A PLS model based on the mid infrared spectra obtained during three replicates of the biotransformation process was constructed. In order to ensure that co-linearity within the system had been adequately addressed the spectral contributors to the model were examined. External validation of the constructed model was achieved by challenging the model with two previously unseen replicates of the process. The constructed model was able to predict the concentrations of two key analytes in various samples from these unseen replicates without the requirement for any time consuming sample pre-treatment stages, thus demonstrating the feasibility of near real-time mid infrared monitoring of such an industrial de-racemization biotransformation process.
Full-text · Article · May 2013 · Analytica chimica acta
[Show abstract][Hide abstract] ABSTRACT: Background
Novel analytical tools, which shorten the long and costly development cycles of biopharmaceuticals are essential. Metabolic flux analysis (MFA) shows great promise in improving our understanding of the metabolism of cell factories in bioreactors, but currently only provides information post-process using conventional off-line methods. MFA combined with real time multianalyte process monitoring techniques provides a valuable platform technology allowing real time insights into metabolic responses of cell factories in bioreactors. This could have a major impact in the bioprocessing industry, ultimately improving product consistency, productivity and shortening development cycles.
This is the first investigation using Near Infrared Spectroscopy (NIRS) in situ combined with metabolic flux modelling which is both a significant challenge and considerable extension of these techniques. We investigated the feasibility of our approach using the industrial workhorse Pichia pastoris in a simplified model system. A parental P. pastoris strain (i.e. which does not synthesize recombinant protein) was used to allow definition of distinct metabolic states focusing solely upon the prediction of intracellular fluxes in central carbon metabolism. Extracellular fluxes were determined using off-line conventional reference methods and on-line NIR predictions (calculated by multivariate analysis using the partial least squares algorithm, PLS). The results showed that the PLS-NIRS models for biomass and glycerol were accurate: correlation coefficients, R2, above 0.90 and the root mean square error of prediction, RMSEP, of 1.17 and 2.90 g/L, respectively. The analytical quality of the NIR models was demonstrated by direct comparison with the standard error of the laboratory (SEL), which showed that performance of the NIR models was suitable for quantifying biomass and glycerol for calculating extracellular metabolite rates and used as independent inputs for the MFA (RMSEP lower than 1.5 × SEL). Furthermore, the results for the MFA from both datasets passed consistency tests performed for each steady state, showing that the precision of on-line NIRS is equivalent to that obtained by the off-line measurements.
The findings of this study show for the first time the potential of NIRS as an input generating for MFA models, contributing to the optimization of cell factory metabolism in real-time.
Full-text · Article · May 2013 · Microbial Cell Factories
[Show abstract][Hide abstract] ABSTRACT: A 785nm diode laser and probe with a 6mm spot size were used to obtain spectra of stationary powders and powders mixing at 50rpm in a high shear convective blender. Two methods of assessing the effect of particle characteristics on the Raman sampling depth for microcrystalline cellulose (Avicel), aspirin or sodium nitrate were compared: (i) the information depth, based on the diminishing Raman signal of TiO(2) in a reference plate as the depth of powder prior to the plate was increased, and (ii) the depth at which a sample became infinitely thick, based on the depth of powder at which the Raman signal of the compound became constant. The particle size, shape, density and/or light absorption capability of the compounds were shown to affect the "information" and "infinitely thick" depths of individual compounds. However, when different sized fractions of aspirin were added to Avicel as the main component, the depth values of aspirin were the same and matched that of the Avicel: 1.7mm for the "information" depth and 3.5mm for the "infinitely thick" depth. This latter value was considered to be the minimum Raman sampling depth when monitoring the addition of aspirin to Avicel in the blender. Mixing profiles for aspirin were obtained non-invasively through the glass wall of the vessel and could be used to assess how the aspirin blended into the main component, identify the end point of the mixing process (which varied with the particle size of the aspirin), and determine the concentration of aspirin in real time. The Raman procedure was compared to two other non-invasive monitoring techniques, near infrared (NIR) spectrometry and broadband acoustic emission spectrometry. The features of the mixing profiles generated by the three techniques were similar for addition of aspirin to Avicel. Although Raman was less sensitive than NIR spectrometry, Raman allowed compound specific mixing profiles to be generated by studying the mixing behaviour of an aspirin-aspartame-Avicel mixture.
No preview · Article · Dec 2012 · Journal of pharmaceutical and biomedical analysis
[Show abstract][Hide abstract] ABSTRACT: Despite the existence of various methods to remove cosmic spikes from Raman data, only a few of them are suitable for process Raman spectroscopy. The disadvantages of these algorithms include increased analysis time, low accuracy of spike detection, or reliance on variable parameters that must be chosen by trial and error in each case. We demonstrate a novel approach to detecting cosmic spikes in process Raman data and validate it using a wide range of experimental data. This new method features a multistage spike recognition algorithm that is based on tracking sharp changes of intensity in the time domain. The algorithm effectively distinguishes cosmic spikes from random spectral noise and abrupt variations of Raman peaks, allowing accurate detection of both high and low intensity cosmic spikes. The procedure is free from variable user-defined parameters and operates reliably in a fully automated manner with a wide range of time-series process Raman data sets containing more than 40 to 50 spectra.
No preview · Article · Nov 2012 · Applied Spectroscopy
[Show abstract][Hide abstract] ABSTRACT: A total of 383 tablets of a pharmaceutical product were analyzed by backscatter and transmission Raman spectrometry to determine the concentration of an active pharmaceutical ingredient (API), chlorpheniramine maleate, at the 2% m/m (4 mg) level. As the exact composition of the tablets was unknown, external calibration samples were prepared from chlorpheniramine maleate and microcrystalline cellulose (Avicel) of different particle size. The API peak at 1594 cm(-1) in the second derivative Raman spectra was used to generate linear calibration models. The API concentration predicted using backscatter Raman measurements was relatively insensitive to the particle size of Avicel. With transmission, however, particle size effects were greater and accurate prediction of the API content was only possible when the photon propagation properties of the calibration and sample tablets were matched. Good agreement was obtained with HPLC analysis when matched calibration tablets were used for both modes. When the calibration and sample tablets are not chemically matched, spectral normalization based on calculation of relative intensities cannot be used to reduce the effects of differences in physical properties. The main conclusion is that although better for whole tablet analysis, transmission Raman is more sensitive to differences in the photon propagation properties of the calibration and sample tablets.
No preview · Article · Apr 2012 · Analytical Chemistry
[Show abstract][Hide abstract] ABSTRACT: Transmission Raman measurements of a 1 mm thick sulfur-containing disk were made at different positions as it was moved through 4 mm of aspirin (150-212 μm) or microcrystalline cellulose (Avicel) of different size ranges (<38, 53-106, and 150-212 μm). The transmission Raman intensity of the sulfur interlayer at 218 cm(-1) was lower when the disk was placed at the top or bottom of the powder bed, compared to positions within the bed and the difference between the sulfur intensity at the outer and inner positions increased with Avicel particle size. Also, the positional intensity difference was smaller for needle-shaped aspirin than for granular Avicel of the same size. The attenuation coefficients for the propagation of the exciting laser and transmitted Raman photons through the individual powders were the same but decreased as the particle size of Avicel increased; also, the attenuation coefficients for propagation through 150-212 μm aspirin were almost half of those through similar sized Avicel particles. The study has demonstrated that particulate size and type affect transmitted Raman intensities and, consequently, such factors need to be considered in the analysis of powders, especially if particle properties vary between the samples.
No preview · Article · Apr 2012 · Analytical Chemistry
[Show abstract][Hide abstract] ABSTRACT: Particle size distribution and compactness have significant confounding effects on Raman signals of powder mixtures, which cannot be effectively modeled or corrected by traditional multivariate linear calibration methods such as partial least-squares (PLS), and therefore greatly deteriorate the predictive abilities of Raman calibration models for powder mixtures. The ability to obtain directly quantitative information from Raman signals of powder mixtures with varying particle size distribution and compactness is, therefore, of considerable interest. In this study, an advanced quantitative Raman calibration model was developed to explicitly account for the confounding effects of particle size distribution and compactness on Raman signals of powder mixtures. Under the theoretical guidance of the proposed Raman calibration model, an advanced dual calibration strategy was adopted to separate the Raman contributions caused by the changes in mass fractions of the constituents in powder mixtures from those induced by the variations in the physical properties of samples, and hence achieve accurate quantitative determination for powder mixture samples. The proposed Raman calibration model was applied to the quantitative analysis of backscatter Raman measurements of a proof-of-concept model system of powder mixtures consisting of barium nitrate and potassium chromate. The average relative prediction error of prediction obtained by the proposed Raman calibration model was less than one-third of the corresponding value of the best performing PLS model for mass fractions of barium nitrate in powder mixtures with variations in particle size distribution, as well as compactness.
No preview · Article · Apr 2012 · Analytical Chemistry
[Show abstract][Hide abstract] ABSTRACT: Analysis of needle-shaped particles of cellobiose octaacetate (COA) obtained from vacuum agitated drying experiments was performed using three particle size analysis techniques: laser diffraction (LD), focused beam reflectance measurements (FBRM) and dynamic image analysis. Comparative measurements were also made for various size fractions of granular particles of microcrystalline cellulose. The study demonstrated that the light scattering particle size methods (LD and FBRM) can be used qualitatively to study the attrition that occurs during drying of needle-shaped particles, however, for full quantitative analysis, image analysis is required. The algorithm used in analysis of LD data assumes the scattering particles are spherical regardless of the actual shape of the particles under evaluation. FBRM measures a chord length distribution (CLD) rather than the particle size distribution (PSD), which in the case of needles is weighted towards the needle width rather than their length. Dynamic image analysis allowed evaluation of the particles based on attributes of the needles such as length (e.g. the maximum Feret diameter) or width (e.g. the minimum Feret diameter) and as such, was the most informative of the techniques for the analysis of attrition that occurred during drying.
[Show abstract][Hide abstract] ABSTRACT: Recently, transmission Raman spectroscopy has been shown to be a valuable tool in the volumetric quantification of pharmaceutical formulations. In this work a Monte Carlo simulation and experimental study are performed to elucidate the dependence of the Raman signal on depth from the viewpoint of probing pharmaceutical tablets and powders in this experimental configuration. The transmission Raman signal is shown to exhibit a moderate bias toward the center of the tablets and this can be considerably reduced by using a recently developed Raman signal-enhancing concept, the "photon diode." The enhancing element not only reduces the bias but also increases the overall Raman signal intensity and consequently improves the signal-to-noise ratio of the measured spectrum. Overall, its implementation with appropriately chosen reflectivity results in a more uniform volumetric sampling across the half of the tablet where the photon diode is used (or across the tablet's entire depth if two photon diodes are used on each side of tablet) and enhanced overall sensitivity. These findings are substantiated experimentally on a segmented tablet by inserting a poly(ethelyne terephthalate) (PET) film doped with TiO(2) at different depths and monitoring its contribution to the overall transmission Raman signal from the segmented tablet. The numerical simulations also indicate considerable sensitivity of the overall Raman signal to the absorption of the sample, which is in line with large migration distances traversed by photons in these measurements. The presence of sample absorption was shown numerically to reduce the signal enhancement effect while the overall depth-dependence profile remained broadly unchanged. The absorption was also shown to produce a depth profile with the photon diode similar to that without it, although with a reduced absolute intensity of Raman signals and diminished enhancement effect.
[Show abstract][Hide abstract] ABSTRACT: Two methods of analysis were developed to permit detection of counterfeit Scotch whisky samples using a novel attenuated total reflectance (ATR) diamond-tipped immersion probe for mid-infrared (MIR) spectrometry. The first method allowed determination of the ethanol concentration (35-45% (v/v)) in situ without dilution of the samples; the results obtained compared well with the supplied concentrations (average relative error of 1.2% and 0.8% for univariate and multivariate partial least squares (PLS) calibration, respectively). The second method involved analysis of dried residues of the whisky samples and caramel solutions on the diamond ATR crystal; principal component analysis (PCA) of the spectra was used to classify the samples and investigate the colorant added. Seventeen test whisky samples were successfully categorised as either authentic or counterfeit in a blind study when both MIR methods were used.
[Show abstract][Hide abstract] ABSTRACT: The evaporation of methanol from needle-shaped particles of cellobiose octaacetate (COA) has been studied directly in a jacketed vacuum drier using in situ measurements by Raman spectrometry. A design of experiments (DoE) approach was used to investigate the effects of three parameters (method of agitation, % solvent loss on drying and jacket temperature), with the intention of minimising the drying time and extent of particle attrition. Drying curves based on Raman signals for methanol and COA in the spectra of the wet particles indicated the end of drying and revealed three stages in the drying process that could be used to monitor the progress of solvent removal in real time. Off-line particle size measurements based on laser diffraction were made to obtain information on the extent of attrition, to compare with the trends revealed by the Raman drying curves. The study demonstrated that non-invasive Raman spectrometry can be used to study the progress of drying during agitation of particles in a vacuum drier, allowing optimisation of operating conditions to minimise attrition and reduce drying times. Although a correlation between particle size and off-line Raman measurements of COA was demonstrated, it was not possible to derive equivalent information from the in situ Raman spectra owing to the greater effects of particle motion or bulk density variations of the particles in the drier.
[Show abstract][Hide abstract] ABSTRACT: Large-scale commercial bioprocesses that manufacture biopharmaceutical products such as monoclonal antibodies generally involve multiple bioreactors operated in parallel. Spectra recorded during in situ monitoring of multiple bioreactors by multiplexed fiber-optic spectroscopies contain not only spectral information of the chemical constituents but also contributions resulting from differences in the optical properties of the probes. Spectra with variations induced by probe differences cannot be efficiently modeled by the commonly used multivariate linear calibration models or effectively removed by popular empirical preprocessing methods. In this study, for the first time, a calibration model is proposed for the analysis of complex spectral data sets arising from multiplexed probes. In the proposed calibration model, the spectral variations introduced by probe differences are explicitly modeled by introducing a multiplicative parameter for each optical probe, and then their detrimental effects are effectively mitigated through a "dual calibration" strategy. The performance of the proposed multiplex calibration model has been tested on two multiplexed spectral data sets (i.e., MIR data of ternary mixtures and NIR data of bioprocesses). Experimental results suggest that the proposed calibration model can effectively mitigate the detrimental effects of probe differences and hence provide much more accurate predictions than commonly used multivariate linear calibration models (such as PLS) with and without empirical data preprocessing methods such as orthogonal signal correction, standard normal variate, or multiplicative signal correction.
No preview · Article · Mar 2011 · Analytical Chemistry