Motivation Investigating Process Behaviour Statistics Background Motivation Investigating Process

Motivation Investigating Process Behaviour Statistics Background Motivation Investigating Process Behaviour Statistics Background Outline 1 CHEE418/801 - Strategies for Process Investigations: Module 1 James McLellan September, 2008 3 2 Motivation Example #1 - Factors Influencing the Formation of Trihalomethanes in Drinking Water Example #2 - Factors Influencing the Grafting of Functional Chemical Groups on Polyethylene Investigating Process Behaviour Why do we need to investigate process behaviour? How do we investigate process behaviour? Why do we need a statistical framework? What do I need to remember from my first statistics course? James McLellan Motivation Investigating Process Behaviour Statistics Background CHEE418/801 Fall 2008 Module 1 Example - THM Formation Example - Melt Grafting James McLellan Motivation Investigating Process Behaviour Statistics Background CHEE418/801 Fall 2008 Module 1 Example - THM Formation Example - Melt Grafting Example #1 - Trihalomethane Formation in Drinking Water Trihalomethanes (THMs) form from chlorination in drinking water treatment plants and water distribution systems THMs are the the most common and dominant chlorinated byproducts in drinking water THMs are probable human carcinogens and may have acute and chronic health effects Human exposure occurs throughout our lifetime via ingestion, inhalation and dermal contact Approximately 14-16% of bladder cancers in Ontario are attributable to drinking water with elevated chlorinated byproducts Can we achieve the necessary disinfection of drinking water while minimizing the formation of chlorinated byproducts? James McLellan CHEE418/801 Fall 2008 Module 1 Thunder Bay Water Treatment Plant Schematic Lake Superior New Reservoirs under construction Ultra Filtration Unit – ZeeWeed 1000 Version 3 New Verticle Turbine High Lift pumps 1350 mm intake Pipe iNTAkE chAMBER PRE-CHLORINATION Bare Point Ultra Filtration Water Treatment Plant iNTAkE & WET WELL Travelling Screens (2) Pre-hypochlorite injection Point Low Lift Pumps (4) PROcESS DiAGRAM FILTRATION RAW WATER cONDUiT SURGE TANkS Ultra Filtration Unit #1 Ultra Filtration Unit #2 Ultra Filtration Unit #3 Ultra Filtration Unit #4 Ultra Filtration Unit #5 POST -CHLORINATION cLEARWELL & RESERVOiR Backwash Pump high Lift Pumps (5) 900 mm intake Pipe EQUALiZATiON TANk 1050 mm Watermain to Distribution system Post-chlorine injection Point To Sanitary Sewer New Plant nearing completion Source: www.thunderbay.ca/docs/water/4060.pdf James McLellan CHEE418/801 Fall 2008 Module 1 Motivation Investigating Process Behaviour Statistics Background Example - THM Formation Example - Melt Grafting Motivation Investigating Process Behaviour Statistics Background Example - THM Formation Example - Melt Grafting The Ontario Drinking Water Surveillance Program As a first step, examine data that are routinely collected to monitor drinking water quality, and see if there are any patterns between the formation of THMs and characteristics of the water source and processing. The Ontario Drinking Water Surveillance Program (DWSP) passively collected data reported by municipal drinking water treatment plants in Ontario 179 different stations reporting - data for 2000-2004 period goal - can we identify factors having an impact on THM formation? factors include Naturally-ocurring Organic Matter (NOM), pH, Chlorine dose, Bromine ion concentration, Temperature, residence time... - 22 possible factors PhD research of Shakhawat Chowdhury James McLellan Motivation Investigating Process Behaviour Statistics Background CHEE418/801 Fall 2008 Module 1 Example - THM Formation Example - Melt Grafting Ontario Drinking Water Surveillance Program Figure: Ontario DWSP Matrix Scatterplot - Ground and Surface Water James McLellan Motivation Investigating Process Behaviour Statistics Background CHEE418/801 Fall 2008 Module 1 Example - THM Formation Example - Melt Grafting Ontario DWSP Data Check An important step in the analysis - see if there are outliers extreme datapoints - that might strongly influence the estimation. THM Data Analysis What are the next steps? correlation and graphical analysis - identify possible trends in the data estimate regression models one challenge - correlation between some of the factors in the database James McLellan CHEE418/801 Fall 2008 Module 1 James McLellan CHEE418/801 Fall 2008 Module 1 Motivation Investigating Process Behaviour Statistics Background Example - THM Formation Example - Melt Grafting Motivation Investigating Process Behaviour Statistics Background Example - THM Formation Example - Melt Grafting Example #2 - Factors Influencing the Grafting of Functional Chemical Groups on Polyethylene Twin-Screw Extruder Has a feed hopper, twin screws in a barrel which is also heated Twin-screw extruder for grafting functional chemical groups on commodity polyethylene Why graft? Improve product value Enhance product workability and applicatons e.g., compatibilization to use for blends, colouring, performance properties (e.g., stiffness) Extruded polymer goes through a water bath for quenching Grafting reactions occur by using an initiator to provide radicals to start reactions - results in grafting of functional chemical groups onto polyethylene backbone, as well as some cross-linking of the polyethylene (undesirable) Performance of the grafting is frequently measured in terms of the grafting efficiency (% of functional group available that is grafted onto polymer) James McLellan Motivation Investigating Process Behaviour Statistics Background CHEE418/801 Fall 2008 Module 1 Example - THM Formation Example - Melt Grafting James McLellan Motivation Investigating Process Behaviour Statistics Background CHEE418/801 Fall 2008 Module 1 Example - THM Formation Example - Melt Grafting Twin-Screw Extruder Grafting Process Improvement Can we find a better set of operating conditions to improve grafting efficiency? Figure: Extruder Hopper improving the extruder grafting efficiency Brainstorming exercise What are the potential factors that have an impact on grafting efficiency? What graphical tool should we use to organize our brainstorming? (Hint- think back to CHEE209 - it’s a spiny thing) Figure: Twin-Screw Extruder Figure: Extruder Barrel with Thermocouples James McLellan CHEE418/801 Fall 2008 Module 1 James McLellan CHEE418/801 Fall 2008 Module 1 Motivation Investigating Process Behaviour Statistics Background Example - THM Formation Example - Melt Grafting Motivation Investigating Process Behaviour Statistics Background Example - THM Formation Example - Melt Grafting Improving the extruder grafting efficiency Improving the extruder grafting efficiency Run experiments, collect data Once we have identified our factors, what experiments should we run? Layout, range, ... Can we get a clear indication of the effect of each operating variable by making good choices of experimental conditions? Now it’s time to analyze the results Goal - identify those operating variables (factors) having a significant impact on grafting efficiency Graphically Quantitatively How can we characterize the impact of operating variables on grafting efficiency? We want to rank the operating variables from most effective to least, and understand the size of their impact James McLellan Motivation Investigating Process Behaviour Statistics Background CHEE418/801 Fall 2008 Module 1 Example - THM Formation Example - Melt Grafting James McLellan Motivation Investigating Process Behaviour Statistics Background CHEE418/801 Fall 2008 Module 1 Why? How? Why statistics? How much variability is there in our experiments? Why do we need to investigate process behaviour? One motivation - to improve ”quality” - of things, processes, life What is Quality? Can we tell this from our experimental results? How will this variability carry through into our calculations? How can we factor in the variability into our decisions about which operating variables are important? American Society for Quality Control - totality of features and characteristics of a product or service that bear on its ability to satisfy given needs Deming (famous quality statistician) - uniformity about target Juran (another famous quality statistician) - fitness for use Taguchi (a more recent, famous quality engineer) - loss imparted to society from the time a product is shipped James McLellan CHEE418/801 Fall 2008 Module 1 James McLellan CHEE418/801 Fall 2008 Module 1 Motivation Investigating Process Behaviour Statistics Background Why? How? Why statistics? Motivation Investigating Process Behaviour Statistics Background Why? How? Why statistics? Quality - Key concepts from definitions product ”delivered” to society - product or process - sold or provided properties attributes of the product typically related to function of product can also be ”passive” properties Implicit Characteristics in Assessing Quality product performance product cost product life / reliability customer satisfaction timeliness environmental impact societal impact and others... target cost specification against which property is judged property ”measurement” from failure to meet specification to producer, society - broad impact of quality How does the THM formation example fit with these definitions of quality? How does the twin screw extruder example fit with these definitions of quality? James McLellan Motivation Investigating Process Behaviour Statistics Background CHEE418/801 Fall 2008 Module 1 Why? How? Why statistics? James McLellan Motivation Investigating Process Behaviour Statistics Background CHEE418/801 Fall 2008 Module 1 Why? How? Why statistics? Quality Improvement Quality Improvement also referred to as ”continuous improvement” in addition to monitoring current quality, strive to improve the quality of the process operation and product concerted effort involving all participants in the organization operators, management, supervisors, unit engineers, designers, administrative staff, ... ... requires knowledge of the effects of factors influencing process and product performance Examples yield vs. temperature, pressure, in a chemical reactor coating uniformity vs. polymer characteristics in a thin film application transmission of variance in two different application methods James McLellan CHEE418/801 Fall 2008 Module 1 James McLellan CHEE418/801 Fall 2008 Module 1 Motivation Investigating Process Behaviour Statistics Background Why? How? Why statistics? Motivation Investigating Process Behaviour Statistics Background Why? How? Why statistics? From an engineering perspective... We try to improve quality by... DESIGN application of fundamental and empirical knowledge to achieve desired objective(s) using, in part, EMPIRICAL KNOWLEDGE empirical = observed from experimentation Some specific requirements for experimental investigation EXPERIMENTATION WHAT should we observe? HOW should we perturb the process? ANALYSIS what happened? SUMMARIZING AND PROJECTION is there a general systematic relationship? summarize using mathematical relationships use these relationships to propose and assess improvements James McLellan Motivation Investigating Process Behaviour Statistics Background CHEE418/801 Fall 2008 Module 1 Why? How? Why statistics? James McLellan Motivation Investigating Process Behaviour Statistics Background CHEE418/801 Fall 2008 Module 1 Why? How? Why statistics? Process Investigations The Iterative Nature of Process Investigations James McLellan CHEE418/801 Fall 2008 Module 1 James McLellan CHEE418/801 Fall 2008 Module 1 Motivation Investigating Process Behaviour Statistics Background Why? How? Why statistics? Motivation Investigating Process Behaviour Statistics Background Why? How? Why statistics? Relationship to this course Why do we need a statistical framework anyway? Processes are inherently subject to variation, which appears in measurements and propagates through any computations made with the measurements. We need a framework which: can characterize this variability can accommodate variability in analysis and computations James McLellan Motivation Investigating Process Behaviour Statistics Background CHEE418/801 Fall 2008 Module 1 Why? How? Why statistics? James McLellan Motivation Investigating Process Behaviour Statistics Background CHEE418/801 Fall 2008 Module 1 Why? How? Why statistics? Where does variability come from? Components in data deterministic process disturbances - internal, external operator interventions instrumentation and measurements how we view the process - approximation as a ”well-mixed” process non-random relationships physical relationships - e.g., material/energy balance relationships in distillation column influence composition in overhead stream stochastic random fluctuations - variability pattern frequently disturbances can evolve in time James McLellan CHEE418/801 Fall 2008 Module 1 James McLellan CHEE418/801 Fall 2008 Module 1 Motivation Investigating Process Behaviour Statistics Background Why? How? Why statistics? Motivation Investigating Process Behaviour Statistics Background What are the roles for statistical methods? What do I need to remember from my first statistics course? Characterizing Variability Patterns random variables probability distributions - density and distribution functions expected values variance, standard deviation, mean graphical methods - histograms, box plots, normal probability plots, ... the Normal probability distribution, Chi-squared distribution, Student’s t distribution, F distribution statistical independence covariance and correlation (we’ll discuss this more in this course) 1 Decision making in the presence of uncertainty (use of confidence intervals, hypothesis tests, comparative box plots) Characterizing variability patterns (e.g., mean, variance) Basis for variability ”accounting” - transmission of variability through calculations Effective presentation of results (graphically, quantitatively) 2 3 4 James McLellan Motivation Investigating Process Behaviour Statistics Background CHEE418/801 Fall 2008 Module 1 James McLellan CHEE418/801 Fall 2008 Module 1 And more... Estimating Characteristics from Data notion of a statistic with associated probability distribution (the sampling distribution) sample average, sample variance and standard deviation degrees of freedom of a statistic Making Inferences hypothesis tests confidence intervals comparison against reference distributions - which one to use? Chi-squared, Normal, Student’s t, F? inferences about means (e.g., single, paired t-tests), variances James McLellan CHEE418/801 Fall 2008 Module 1