Presentation on theme: "Process Analytics: Improving Measurement Capability in your Plant AIChE Meeting: Nov. 17, 2009 Steve Wright Process & Environmental Analytics Eastman Chemical."— Presentation transcript:
Process Analytics: Improving Measurement Capability in your Plant AIChE Meeting: Nov. 17, 2009 Steve Wright Process & Environmental Analytics Eastman Chemical Company
Overview of Presentation Introduction Sampling Theory – Why Analyzers? Process Analyzers and Sensors Sampling Systems Ownership and Maintenance Summary and Q&A
Introduction Process Analytics is the Analytical Measurement of Chemical Composition Chemical Properties of a Chemical Production Stream Using one, or more, of four approaches In-Situ / In-Line Extractive At-Line Ambient Detection 100 ppm 100 ppm
Eastman Process Analytics: Kingsport Almost 2000 Process Analyzers and Chemical Sensors Personnel: 41 chemists, engineers, techs and analysts. Support: 24x7 where needed. Responsibilities include: Analyzer consultation/ specification Analyzer system design/purchase Sample system construction Installation/checkout Preventative maintenance Reactive maintenance Analyzer succession planning
Measurement & Control MEASUREMENT CONTROL Can’t do one without the other..
Majority of Process Measurement Tasks Can be Done Using Simple Sensors or Lab Analyses Pressure, Temperature, Flow, Mass Lab measurements – slow and steady processes.. For the exceptions – process analytics..
Traditional Reactor Sampling Path Insert Break Lab Queue Enter Sample Order Ye Olde Bucket/Spigot Wait for Truck Wait for Truck II GC Results Sample Point Create Report Results! Insert Lunch
SP Time, Minutes + - SP + - Time, Minutes GARBAGE ZONE Under-Sampling (Aliasing)
Comfort Zone – >4F to “over-sampling” SP Time, Minutes Meeting Nyquist.. Just barely + - SP + - Time, Minutes GARBAGE ZONE Sampling Okay – but it’s breaking the bank.. LAB $$$$$$$$$$$$$$$$$$$
Transitioning from Product A to Product B How long can we continue making “Good A” And when can we call “Good B”? Good Measurement can Lead to Great Processes 100% 0% ? Good A Good B ?? Time
Target Time Late Early Early Error Lead Late Error Lag Time Process Variable “While we’re here, let’s save time & take the next sample” “Just trying to keep up” Input Lead/Lag Control Complexities…. Variability Complicates it even More… Crew 1 is timely.. Crew 2 isn’t.. Process trend Sample Time-Stamping Errors Early Late
Good Measurement can Lead to Great Processes A B at 70 deg C 0-100% in 20 minutes A B at 20 deg C 0-100% in 640 minutes While waiting 2 hours for analysis 20% change!!! Two Hours 70 deg C 20 deg C % Completion 60% 80% The Reaction Continues…
Models Ultimate process understanding “victory” Control process with lean measurements, T, P, flow, etc. “As good as their input data” Model response surface must be well-defined Models tend to perform best in “known territory” Prediction weakness can occur during critical times: Upsets Start-Ups/Shut-Downs Product transitions Direct measurement benefits Often easier to set up and maintain than complex models Full process interaction / understanding is not required to measure Output can help build better models!
“The Wall” Loose measurement & control – broad process performance Tight measurement & control “Wall” = > Impurities, lower value product, permits, safety issues, etc. Run Closer… Value Proposition
Process Analyzers & Sensors
Fixed or Dedicated Systems Transmitter Style o Simple to install o Low cost o Can be used as “cheap” analyzers - If high accuracy not required - Inferential o Extractive sampling of processes - Addition of flow cells and sample system o Direct Insertion Alan Hensley, 2009
Ambient / Area Point Monitoring Personnel Protection Leaks or Spills Electrochemical o Toxic Gases Hydrogen sulfide Chlorine dioxide Carbon monoxide o Oxygen % levels Oxygen deficiency High oxygen in processes Alan Hensley, 2009
Area / Point Monitoring Combustible Gases o Normally report values in terms of % of lower explosive limit (LEL) o Not specific to gas – detect hydrocarbons o Catalytic sensor Combust the sample Require oxygen o Infrared sensor Can be used in oxygen deficient or inert environments Where “poisoning” of catalytic sensor is of concern Alan Hensley, 2009
Liquid Analytical pH & Conductivity o Sumps / Pure Water / Condensate & Discharges Material release Quality Contamination o Process Inferential composition measurement pH control for reactions / batch processes Dissolved Oxygen o Wastewater Treatment
Oxygen Fuel Cell % and ppm level measurements Paramagnetic o % level measurements o Oxygen level in nitrogen convey systems Zirconium Oxide o Stack monitoring o Handle dirty environments o High temperature operation Alan Hensley, 2009
Physical Property Density (not just for mass flow) o Inferential Composition Measurement o Depends on process stream Turbidity o Contamination Viscosity o Process Control Alan Hensley, 2009
Fixed or Dedicated Systems Traditional Style o Analyzer is remote from area o Extractive sampling with sample system o Higher cost o Higher complexity o Require more care and feeding o Stream Complexity o Accuracy o Specific
Photometric Methods Photometers UV/NIR/Non-dispersive IR Use specific wavelengths = 1 or 2 components Solids: non-contact o % moisture Gases o CO, CO2, NOx, SO2 Liquids o % water, % organic acids Alan Hensley, 2009
Process Analyzer Availability If it can be done in the lab, it can installed on-line. Up-Front cost issues / ROI Sample handling issues Our group will build it if we cannot buy it. Integration tasks, sample handling systems Panel Shop in B-359A FRONT VIEW
Sampling Systems Maintenance
Goals Representative Sampling / Minimal Handling Want sample to mirror process content Minimal interaction with sample Minimal sampling delays Sample Compatibility with Analyzer Specifications Temperature Pressure Flow Viscosity Particulates / Bubbles Materials Compatibility
In-situ Measurement No sampling system Pipeline/tank/line insertion No delay – real time results Probe design Probe can be removed for cleaning – usually. Exceptions would be high pressure / temperature applications Representative! Passive? Yes.
Extractive Sampling Systems Sample stream removed from main process line Advantages Allows isolation from process (cleaning/calibration) Filtration, dilution, P/T manipulation Improved safety – block/bleed Difficulties Must have dP across sample loop - or Sample pumping Delays Returning altered material to process or waste Filtration maintenance – when needed
Process Analyzer Maintenance Ownership Cycle..
Development Purchase Installation Start-Up Maintenance Improvement Replacement 70-90% Cost of Ownership Maintenance is major cost of analyzer installation – process GC example
Reliability Maintenance Approach Reliability – in degrees.. Ideal Lasts forever, accurate and precise – cheap to own too. Reality (March to Entropy) Machine components wear out Unusual, unexpected events happen Goal Want capable function whenever machine is needed High Availability Uptime. Want ownership at lowest possible cost. Reliability-Centered Maintenance Approach
Maintenance Categories Reactive (RTF) Appropriate for ultra-high reliability, low criticality systems Cheapest / Most Expensive approach – feeling lucky? Preventative (PrM) Process analytics use PrM Shewhart control charts (+/- 3 sigma, run of eight) Scheduled benchmarking visits Predictive (PdM) Maximum system availability at minimum cost. Relies on obtaining detailed history at component levels. We now have tools in-place to transition to PdM where needed.
pH 4 pH X.XX pH 3.95 pH 4.15 Analyzer Benchmarking Apply standard of known concentration to analyzer Read analyzer response Compare response to standard response Within control limits: Note response, add to control chart Walk away, just walk away….. Outside of control limits, or eight either side of average: Note response, add to control chart Calibrate analyzer / determine cause / log Avoids human tendency to over- control, chasing system noise. Much better for process stability. pH 4.0 pH 3.9 pH -3 Target pH 4.00 pH 4 Standard Buffer Solution
Process Analyzer Maintenance Effective PrM has greatly improved reliability of our analyzers High availability up-times Analyzer data can be trusted for monitoring & control Productivity (analyzers/analyst) has greatly improved over the last 20 years Better analyzer technology Better diagnostics Scheduled PrM Improved tools such as OSI PI
Questions & Answers Thanks! Steve Wright Senior Development Associate Process and Environmental Analytics Bldg 359A Eastman Chemical Company Phone: