Process analytics in PI plant – the challenges David Littlejohn University of Strathclyde/CPACT PROCESS INTENSIFICATION: Meeting the Business and Technical Challenges, Gaining Competitive Advantage The Royal Institution of Great Britain, London: 19th November 2003 Reasons for process analysis • • • • • • • Monitor progress of a reaction Know when end-point is reached Check reaction kinetics Impurity detection Monitor/control blending or mixing Used for decision making or direct control Requires timely qualitative or quantitative analysis Increased profits through on-line process analysis & control • • • • • • Better yield of product Greater purity of product More consistent production Less re-work Lower energy usage Improved use of raw materials etc. etc. Terminology Off-line - At-line - On-line - Manual sampling and analysis at a separate laboratory Manual sampling and analysis at the process Automatic sampling and analysis; process - analyser interface On-line analysis True on-line - In-line - Non-invasive - Some of the process liquors are passed through an on-line analyser No sampling; analyser probe inserted into reactor or pipe No sampling; no direct contact required between process liquors and analyser Features of process analysis in conventional plant • Kit is comparatively large • Probes in sampling loops or insertion devices • Reaction/operation times are 10s min to hours • Measurement times are seconds to minutes • Measurements repeated during process • Calibration/validation measurements included Process analysis • Spectroscopy techniques - NIR, mid-IR, UV-visible, Raman, MS, NMR, XRF • Gas chromatography and mass spectrometry • Fast response times • Information rich • Chemical composition monitoring • From process development to production Checklist for implementation of process analysis techniques • • • • • • Decide information required Decide frequency Consider sampling issues Select technique Decide measurement technology Consider calibration/modelling Schematic of a dye manufacturing process Monitoring dye process by visible spectrometry • Dye content - 200 g/l • Effluent - 0.1 g/l • Optimum path lengths are 0.00025 and 0.5 cm, respectively. • Use “transmission” for effluent and “ATR” for concentrate Transmission Probe 2 1 Attenuated Total Reflection Probe 3 Experimental arrangement for attenuated total reflection Challenges in signal processing and calibration • Usually measure “mixed” spectra • Spectra often broad-band • Some parts of spectra are more useful than others • Samples change composition with time • Lab-prepared standards may not be representative Features of process analysis in PI plant • Kit is small and may be a continuous process • Reaction/operation times are seconds to minutes • Access for probes may be limited • Measurement times must be short e.g. seconds • Need robust transferable calibration procedure or methods that are calibration-free • Forward control of rapid multi-step processes may be desirable Innovative Reactors Innovative Technologies Static Mixers Innovative Technologies FlexReactor from BHR Group FlexReactor - Features Features Simple but effective static mixer technology Highly flexible package Wide range of materials of construction • Benefits – Use for wide range of processes – Flexibility to cope with undefined chemistry Measurement characteristics in PI • In-situ measurements better than extractive sampling • Can flow-cells be built-in to continuous PI processes? • Non-invasive measurements avoid disturbing process dynamics Features of techniques for PI Gas chromatography • • • • • Possible for gaseous processes Extractive sampling required Good sensitivity range Compact systems now available Fast analysis possible Micro machined GC system analytical module (c) Siemens Micro machined GC system - module assembled to basic unit (c) Siemens Multi- and in-line detection (c) Siemens Laboratory GC Injector Conventional Process GC Injector Column Detector Column Column Backflush Column Distribution Detector Column Distribution Column Column Process-GC with Parallel Chromatography Approach Injector Column Backflush Column Detector Features of techniques for PI Mass spectrometry • Extractive sampling required • Rapid and sensitive analysis possible • Requires chemometrics to deconvolute fragmentation patterns • Improvements to design required for PI applications? Features of techniques for PI • • • • • • • • NIR spectrometry Not particularly sensitive Fast measurement times <1 s, but….. Used for liquids and powders Flow-cells possible Compatible with silica fibres Chemometric calibration required Transmission and reflection options Non-invasive use possible (needs window) Zeiss Corona 45 NIR Aspect software used to acquire spectra through PC link Powder blending glass mixing vessel transducer attached here Zeiss Corona (non-invasive NIR spectrometer) Features of techniques for PI • • • • • • • Raman scattering Fundamentally a fast event Not a sensitive technique Measurement times may be too long Fluorescence can be a problem Compatible with silica fibres Can be non-invasive Sharper spectra than NIR Kaiser HoloProbe Raman Spectrometer Non-invasive Raman • Non-contact optic attached to probe head • Probe to bottle distance = 5.5 cm • Measurement time = 3 min Features of techniques for PI • • • • • • Mid-infrared spectrometry Can be relatively sensitive Good at molecular identification Flow-cell for gases, ATR probe for liquids Not easy to interface with plant Special materials/fibres required Multivariate or univariate calibration SpectraProbe Linx 5-10 Acoustic measurements • Passive transducer • Active pulse-echo pitch-catch Acoustic impedance incident transmitted reflected material 1 impedance Z1 material 2 impedance Z2 • If Z1 and Z2 are very different (e.g. air and water) most of the incident wave will be reflected • If Z1 and Z2 are similar (e.g. oil and water) most of the incident wave will be transmitted Impact on measurements passive – yes active - no e.g. granular solid material passive – yes active - yes e.g. solid in liquid passive – no active - no glass air glass vessel wall passive – yes active - yes glass oil glass vessel wall Acoustic emission • Can measure and process the data to give spectrum that is based on “frequency of event” or acoustic frequency • Non-invasive and relatively inexpensive • Responds to changes in physical form • Not a molecular technique • Not particularly sensitive • Requires sophisticated signal processing Acoustic emission event frequency Plant Sensor and preamplifier Data acquisition and prediction DC signal envelope Amplified AC signal AC to RMS conversion log10(intensity/a.u.) Event frequency spectrum 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 Stirring of heterogeneous mixture 0 5 10 15 20 frequency of event/Hz 25 30 Nano 30 broadband transducer (150-750 kHz) pre-amplifier and filters transducer power supply oscilloscope GPIB/USB computer Matlab Broadband transducer Linked directly to oscilloscope Agilent Infiniium oscilloscope Data acquired in C-program using GPIB link Signal processing Fourier transform process time time sum frequency frequency Passive acoustic spectrum of aspirin in avicel – Nano 30 0.14 intensity/a.u. 0.12 0.1 0.08 0.06 0.04 0.02 0 50 150 250 frequency/kHz 350 200 g of itaconic acid in toluene Stir rate = 250 rpm Oil temp = 20 °C 6000 intensity/a.u. 5000 4000 3000 2000 1000 0 0 100 200 300 frequency/kHz 400 500 Chip-based approach to LC • Fast analysis times • Micro-machined capillary column – Ideal column geometries realizable – High fabrication uniformity • Integrated injector and detector – No extra-column dispersion • No column packing – low operating pressures Technological challenges • Reproducible sample injection • Interfacing the macro-world with a microfluidic chip • Continous flow sampling • Avoidance of sample carry-over between measurements 2nd generation LC chips © Crystal Vision Microsystems Sample inlet Buffer inlet Sample outlet Electrode contact pads 20mm Outlet Details of chip mount © Crystal Vision Microsystems Conception of analytical module © Crystal Vision Microsystems Analytical module Schematic of module and deployment Some final thoughts • Most current process analysers are not designed for PI plant in size, sampling interface etc. • Techniques that you think could be good chemically may be difficult to implement • Sensitivity limitations could be a problem considering short process times • Combinations of techniques may be required • Curve resolution chemometrics will be important to achieve calibration-free analysis Some final thoughts (continued) • Analysers must be built-in if possible • On-line optimisation and control will be required in many processes New approaches to the design and operation of process analysers are required