Realtime Metrics for Process Performance By focusing on realtime measures, plant operators can improve bottom-line results



Similar documents
Enhancing Oracle Business Intelligence with cubus EV How users of Oracle BI on Essbase cubes can benefit from cubus outperform EV Analytics (cubus EV)

*The most important feature of MRP as compared with ordinary inventory control analysis is its time phasing feature.

HCL Dynamic Spiking Protocol

COMPARISON OF THE EFFICIENCY OF S-CONTROL CHART AND EWMA-S 2 CONTROL CHART FOR THE CHANGES IN A PROCESS

CHAPTER 3 THE TIME VALUE OF MONEY

To c o m p e t e in t o d a y s r e t a i l e n v i r o n m e n t, y o u n e e d a s i n g l e,

CCH Accountants Starter Pack

A guide to School Employees' Well-Being

SECTION 1.5 : SUMMATION NOTATION + WORK WITH SEQUENCES

Agenda. Outsourcing and Globalization in Software Development. Outsourcing. Outsourcing here to stay. Outsourcing Alternatives

iprox sensors iprox inductive sensors iprox programming tools ProxView programming software iprox the world s most versatile proximity sensor

How to read A Mutual Fund shareholder report

Domain 1: Designing a SQL Server Instance and a Database Solution

Forecasting. Forecasting Application. Practical Forecasting. Chapter 7 OVERVIEW KEY CONCEPTS. Chapter 7. Chapter 7

(VCP-310)

Analyzing Longitudinal Data from Complex Surveys Using SUDAAN

Agency Relationship Optimizer

Incremental calculation of weighted mean and variance

Confidence Intervals for One Mean

The Forgotten Middle. research readiness results. Executive Summary

Flood Emergency Response Plan

Bio-Plex Manager Software

France caters to innovative companies and offers the best research tax credit in Europe

IntelliSOURCE Comverge s enterprise software platform provides the foundation for deploying integrated demand management programs.

5: Introduction to Estimation

Measures of Spread and Boxplots Discrete Math, Section 9.4

In nite Sequences. Dr. Philippe B. Laval Kennesaw State University. October 9, 2008

GCSE STATISTICS. 4) How to calculate the range: The difference between the biggest number and the smallest number.

Prescribing costs in primary care

Data Analysis and Statistical Behaviors of Stock Market Fluctuations

Wells Fargo Insurance Services Claim Consulting Capabilities

This document contains a collection of formulas and constants useful for SPC chart construction. It assumes you are already familiar with SPC.

How to use what you OWN to reduce what you OWE

CCH CRM Books Online Software Fee Protection Consultancy Advice Lines CPD Books Online Software Fee Protection Consultancy Advice Lines CPD

A Guide to the Pricing Conventions of SFE Interest Rate Products

Evaluating Model for B2C E- commerce Enterprise Development Based on DEA

CCH Practice Management

CONTROL CHART BASED ON A MULTIPLICATIVE-BINOMIAL DISTRIBUTION

BENEFIT-COST ANALYSIS Financial and Economic Appraisal using Spreadsheets

Online Banking. Internet of Things

Study on the application of the software phase-locked loop in tracking and filtering of pulse signal

Case Study. Normal and t Distributions. Density Plot. Normal Distributions

AGC s SUPERVISORY TRAINING PROGRAM

FIRE PROTECTION SYSTEM INSPECTION, TESTING AND MAINTENANCE PROGRAMS

Rainbow options. A rainbow is an option on a basket that pays in its most common form, a nonequally

1 Computing the Standard Deviation of Sample Means

A GUIDE TO BUILDING SMART BUSINESS CREDIT

Is there employment discrimination against the disabled? Melanie K Jones i. University of Wales, Swansea

Ideate, Inc. Training Solutions to Give you the Leading Edge

Chapter 5 Unit 1. IET 350 Engineering Economics. Learning Objectives Chapter 5. Learning Objectives Unit 1. Annual Amount and Gradient Functions

ANALYTICS. Insights that drive your business

CHAPTER 3 DIGITAL CODING OF SIGNALS

client communication

Automatic Tuning for FOREX Trading System Using Fuzzy Time Series

Z-TEST / Z-STATISTIC: used to test hypotheses about. µ when the population standard deviation is unknown

Investing in Stocks WHAT ARE THE DIFFERENT CLASSIFICATIONS OF STOCKS? WHY INVEST IN STOCKS? CAN YOU LOSE MONEY?

Valuing Firms in Distress

CCH Accounts Production

A Balanced Scorecard

Modified Line Search Method for Global Optimization

Output Analysis (2, Chapters 10 &11 Law)

Comfort for Life CAPT CAPF CHPF CAUF CSCF

SOCIAL MEDIA. Keep the conversations going

CREATIVE MARKETING PROJECT 2016

1 Correlation and Regression Analysis

Optimize your Network. In the Courier, Express and Parcel market ADDING CREDIBILITY

LEASE-PURCHASE DECISION

Lesson 17 Pearson s Correlation Coefficient

Baan Service Master Data Management

facing today s challenges As an accountancy practice, managing relationships with our clients has to be at the heart of everything we do.

Vladimir N. Burkov, Dmitri A. Novikov MODELS AND METHODS OF MULTIPROJECTS MANAGEMENT

Advancement FORUM. CULTIVATING LEADERS IN CASE MANAGEMENT

A Guide to Better Postal Services Procurement. A GUIDE TO better POSTAL SERVICES PROCUREMENT

ODBC. Getting Started With Sage Timberline Office ODBC

IT Support n n support@premierchoiceinternet.com. 30 Day FREE Trial. IT Support from 8p/user

HOSPITAL NURSE STAFFING SURVEY

InventoryControl. The Complete Inventory Tracking Solution for Small Businesses

Chatpun Khamyat Department of Industrial Engineering, Kasetsart University, Bangkok, Thailand

5.4 Amortization. Question 1: How do you find the present value of an annuity? Question 2: How is a loan amortized?

UM USER SATISFACTION SURVEY Final Report. September 2, Prepared by. ers e-research & Solutions (Macau)

Non-life insurance mathematics. Nils F. Haavardsson, University of Oslo and DNB Skadeforsikring

Ken blanchard college of business

Sequences and Series

Bond Valuation I. What is a bond? Cash Flows of A Typical Bond. Bond Valuation. Coupon Rate and Current Yield. Cash Flows of A Typical Bond

Hypothesis testing. Null and alternative hypotheses

RECRUITMENT TRENDS SURVEY RESULTS

The analysis of the Cournot oligopoly model considering the subjective motive in the strategy selection

Erik Ottosson & Fredrik Weissenrieder, CVA. Cash Value Added - a new method for measuring financial performance.

A GUIDE TO LEVEL 3 VALUE ADDED IN 2013 SCHOOL AND COLLEGE PERFORMANCE TABLES

Soving Recurrence Relations

Exam 3. Instructor: Cynthia Rudin TA: Dimitrios Bisias. November 22, 2011

INVESTMENT PERFORMANCE COUNCIL (IPC)

Hypergeometric Distributions

FM4 CREDIT AND BORROWING

Systems Design Project: Indoor Location of Wireless Devices

Domain 1 - Describe Cisco VoIP Implementations

Full Lifecycle Project Cost Controls

Lesson 15 ANOVA (analysis of variance)

Statement of cash flows

Professional Networking

Transcription:

Feature Egieerig Report Practice Realtime Metrics for Process Performace By focusig o realtime measures, plat operators ca improve bottom-lie results George Buckbee ExperTue, Ic. Process performace, quality ad reliability deped directly o how a plat performs i real time. Yet may plats track ad respod oly to logterm averages. By measurig ad respodig to realtime measures, you ca esure: Quick recovery from process upsets Fast chage-over to aother product Shorteed batch-cycle times Measurig ad respodig to the right dyamic measures will give you isight ito the process, ad help you to drive plat improvemets. Plats that use realtime metrics perform substatially better tha plats that do ot. Accordig to the recetly-released Metrics That Matter study [1], those who use plat dashboards were 37% more likely tha others to improve agaist operatios' key performace idicators (KPIs). These plats were also 53% more likely to improve bottomlie busiess metrics by more tha 1%. Criteria for realtime measures May plats ted to focus o statistical averages of performace, such as average cost, percet rejects, or percet uptime. These are excellet measures of process results, but i order to make improvemets, we eed to focus o realtime measures of performace. To be useful, these FIGURE 1. measures should meet the followig three criteria: 1. Meaigful. Measurable 3. Actioable Whe a measure is actioable, it helps to lead you directly to a corrective actio. The key to makig process improvemets is to measure the right thigs, ad the to respod with the right corrective actios. Some of the most commo realtime measures that meet this criteria are preseted i this article, alog with a discussio o how these measures ca be made available via a performacesupervisio system (PSS). A PSS moitors realtime data, calculates performace metrics, presets the iformatio usig plat dashboards, ad provides the tools eeded to make performace improvemets. How to get realtime data With the advet of moder distributed cotrol systems (DSCs) ad ope commuicatios techologies, such as OLE for Process Cotrol (OPC), today s process plat has access to a wealth of realtime iformatio. A realtime PSS ca cruch through all the raw data ad develop meaigful realtime performace measures, the most importat of which are described below. With a DCS or PLC cotrol system, the process plat is already gatherig realtime data from istrumets throughout the plat. I fact, cotrol systems have bee doig this for years. More recetly, cotrol-system vedors have opeed up their proprietary systems. OPC is a series of commuicatios stadards that provide ope coectivity i idus- 54

trial automatio systems. (See glossary, top right) Oce the data is available via OPC, it ca be shared with: Other cotrol systems Supervisory cotrols Realtime data historias Performace-supervisio systems Performace-supervisio systems, i fact, will perform may of these calculatios automatically, based o the available realtime data. Figure 1 shows how realtime data from plat istrumetatio ca be fed via OPC or historias ito a PSS. For these systems, web-based plat dashboards ca make the data accessible to users throughout the plat or eve the corporatio. The remaider of this article focuses o how to calculate a umber of realtime performace measures. Basic measures Stadard deviatio. The stadard deviatio is a basic statistical measure of process performace. It is calculated over a period of time, from all the available data i that time period (x i ) ad the average value of the data (x ), as show i Equatio (1). i1 xix 1 FIGURE. FIGURE 3. (1) The stadard deviatio ca provide some isight ito the variability of the process. For example, tredig stadard deviatio over time will help to idetify chages i process variatio. Stadard deviatio is ofte used to develop some iterestig performace measures, as will be discussed below. Variace (regular ad ormalized). Variace (V) is defied as the square of the stadard deviatio (): V () Stadard deviatio ad variace deped, of course, o the egieerig uits of the raw measuremet. It would ot make sese, for example, to compare the variace of a temperature measuremet with that of a tak level. I fact, it ca eve be challegig to iterpret the value of the regular variace. I a cotrol system, realtime process measuremets are boud by both upper- ad lower-rage limits. The upper ad lower process variables (PV) ca be used to covert variace to a ormalized versio (V ormal ), as show i Equatio (3). V ormal GLOSSARY CO Cotroller output DCS Distributed cotrol system OLE Object likig & embeddig: A Microsoft stadard for commuicatio betwee applicatios. OPC OLE for Process Cotrol: A series of stadards specificatios for ope coectivity i idustrial automatio PLC Programmable logic cotroller PSS Performace supervisio system PV Process variable SP Set poit V(1) PV PV hi lo (3) Normalized variace ca be compared betwee sesors i the plat. I fact, if you sort all sesors i the plat accordig to ormalized variace, you are likely to immediately fid some poorly-performig sesors: Sesors with the highest ormalized variace may idicate istallatio errors, poor wirig termiatio, or excessive maipulatio of the cotrol loop Sesors with very low ormalized variace may be flat-lied, idicatig a loss of commuicatios, uitetioal removal from service, or a failed sesor Process performace Beyod the basic statistics, meaigful process iformatio ca be gathered from realtime data. P p, ad P pk. P p ad P pk are statistical measures of process capability. I effect, P p ad P pk measure the ratio of the process specificatios to the demostrated capability of the pro- 55

Egieerig Practice cess. Process-specificatio limits ca be give as USL (upper specificatio limit) ad LSL (lower specificatio limit). Equatio (4) demostrates the calculatio of P p. P p USL LSL 6 (4) P pk is a better measure whe the process is ot cetered betwee the specificatio limits. Equatios (5 7) demostrate how to calculate P pk. P pl P p pk x LSL 3 (5) USL x (6) 3 pl pu P Mi P, P (7) Whe P p ad P pk are greater tha 1, the process is capable of meetig the specificatio. Values less tha 1 idicate that the process has so much variability that it caot be expected to meet specificatios. Opportuity gap The opportuity gap takes P p ad P pk oe step further. I most processes, there is a ecoomic icetive to shift process operatios toward oe of the specificatio limits. I dryig systems, for example, eergy is saved by operatig as close to the upper moisture specificatio as possible. This is sometimes referred to as crowdig the spec, ad it is oe of the fastest ways to save moey i a process plat. Opportuity gap provides a actioable measure. Basically, it recommeds a setpoit adjustmet that will move the process closer to the spec limit. Figure shows how the opportuity gap is calculated. Savigs with opportuity gap. Opportuity gap saves moey directly. Whe the operator closes the opportuity gap, the plat starts savig moey istatly. As a practical matter, it is usually simple to calculate the savigs potetial from opportuity gap. For the most importat process quality variables, operatig departmets are well aware of the value of icremetal improvemet. For example, a paper mill will likely kow the value of a 1% moisture shift. A example of this calculatio is show below, for a typical dryig operatio. $883/ hour SavigsRate % Moisture (8) hours Savigs/ day OppGapSavigsRate4 day (9) With a Opportuity Gap of oly.5% moisture, makig the setpoit chage is worth over $1,/d, i reduced eergy costs. Oscillatio detectio Oscillatios of a process variable create may problems i process plats. They propagate throughout the plat, icreasig variability, icreasig eergy costs, ad destabilizig may operatios. I a recet study, we have foud that it is typical for 4% of cotrol loops to be oscillatig. Detectig ad elimiatig oscillatios i real time will help to drive the plat to its peak performace. Imagie the differece betwee drivig a car dow the highway at a steady speed, versus drivig while cyclig pressure o the accelerator. While the car may average the same speed, its fuel efficiecy is much better without the oscillatio. FIGURE 4. Fourier aalysis fids the cycles. Accordig to Fourier s theory, ay periodic sigal ca be broke dow ito its compoet frequecies, as show i Equatio (1). f a t a cost b si t 1 (1) The idividual coefficiets a ad b ca be foud from Equatios (11) ad (1) a f tcos tdt tsi tdt (11) b f (1) Luckily, moder PSS systems ca cruch through these umbers with ease. Simply look for the largest coefficiets to fid the highest frequecies. Sources of oscillatio. All loops that oscillate at the same frequecy are probably oscillatig due to a commo cause. Sort all the loops i your plat by oscillatio period, ad you will quickly see which oes are iteractig, ad which are ot. Start upstream i the process. Look 56

t at 63%rise (16) Higher-order models require more complex approaches. The moder approach, usig techiques such as Active-Model-Capture techology, uses software to automatically capture process-bump data, determie the form of the model, calculate the model parameters, ad validate the results. for loops that oscillate at the same frequecy i the upstream process. It is very commo to see a boiler loop, for example, drivig a oscillatio through a etire refiery. Aother way to isolate the root cause of a oscillatio is to ope the cotrol loop. If the oscillatio disappears whe the loop is put i maual, the you have foud the root cause. If the oscillatio remais whe the loop is i maual, the you should cotiue lookig upstream. Cotroller performace There are coutless ways to measure the performace of process cotrols. Some of the most commo performace metrics are show i this sectio. Itegral of absolute error (IAE). This classical measure of cotrol performace will tell you how well the process variable tracks to the setpoit. This calculatio is quite useful for comparig the results of differet cotrol strategies o a sigle cotrol loop. The basic relatio for IAE is show i Equatio (13). IAE SP PV dt (13) FIGURE 5. To be practical, we ca t itegrate to ifiity, so we typically itegrate over a fixed period of time for comparisos. IAE is affected by may other factors, such as umber ad size of setpoit chages, ad load upsets. Whe comparig IAE or other cotrol performace measures, you should be sure to compare uder very similar coditios. Harris idex. The Harris idex, I H, is used to compare the performace of your cotroller ( act ) to the best possible feedback cotrol, otherwise kow as Miimum-Variace cotrol. Calculatio of the Harris idex is show i Equatio (14) I H act MiVar (14) A Harris idex close to 1 meas the cotroller is performig very well, ad a large Harris idex idicates opportuity to improve the cotrol. Dyamic process models I the past, process models were calculated usig offlie graphical techiques. Figure 3 shows a graphical techique for calculatig a first-order, plus time delay model. Three model parameters are calculated: dead time (t d ), process gai (G p ) ad time costat (). To use the graphical techique, gather process-variable ad cotroller-output (CO) data from your cotrol system, usig a step test. The deadtime is determied by ispectio. For a first-order, plus time delay system, calculate the values for process gai ad dead time as show i Equatios (15) ad (16). PV G p CO (15) Frequecy-domai models For eve more capability, a frequecy-domai model ca be geerated. This will allow you, for example, to idetify resoat frequecies for the closed-loop system. This iformatio ca help to coordiate resposes amog cotrollers. For example, coordiated resposes are required to esure the success of cascade, ratio, ad feedforward loops. A example of a frequecy-domai model is show i Figure 4. Robustess Robustess is a measure of how well a cotrol loop will respod uder chagig process coditios. Oe way to measure robustess is usig a robustess plot, as show i Figure 5. I a robustess plot, the cotroller tuig is represeted by a lie o the chart. The process model is show by the cross-hair at the bottom left. Differet sets of cotroller tuig will draw differet lies. As the tuig lie approaches the cross-hairs, the cotrol becomes oscillatory, ad evetually will become ustable. Whe model dyamics are automatically determied by Active Model Capture, robustess calculatios ca be doe i real time. Cotrol egieers ca be immediately otified of ay stability or cotrol issues. Expertise built ito metrics With the tred of reduced iteral persoel, there is a eed for more ad more expertise to be built ito automated systems. A moder PSS will icorporate dozes of years of egieerig expertise ito its software. With this embedded expertise, juior egieers icrease their capability. Seior egieers quickly drill ito the critical process iformatio, 57

Egieerig Practice greatly reducig the time required for data collectio, aalysis ad review. This helps to keep egieers ad techicias at all levels focused o the value-added work of performace improvemet. Metrics deliver results Process improvemet, variability reductio, ad savigs go had-i-had. This has bee prove time ad agai throughout the process idustries. A few examples of this iclude: Kruger Paper saved over $1,, usig a PSS for its realtime diagostics durig a ew process startup Columbia Chemicals, producer of chemicals such as carbo black, has saved over $1,, i eergy costs. Columbia tracked realtime process performace to reduce product variability, save eergy ad icrease productio rates A U.S. cemet plat saw a 16% productio icrease by elimiatig cyclig problems A oil refiery i Asia improved aual profits by millios of dollars. Usig the right metrics for robustess was key to their success Accordig to Larry Spickard, seior vice presidet of maufacturig at Columbia Chemicals, The results ad beefits are very importat to our busiess. These are part of stadard key operatig metrics that we track at the seior maagemet team of the compay. They directly relate to our fiacials, ad are really just part of the overall performace that we brig to our customers. Fial remarks Realtime metrics deliver substatial beefits to process plats. Meaigful, measurable, ad actioable metrics are readily available i real time, usig a performace supervisio system. These performace metrics ca be used to ucover plat performace problems, to improve process cotrol, ad to drive cotiuous improvemet. Usig these metrics, plats ca focus efforts ad quickly get to the root cause of performace issues. Edited by Gerald Odrey Refereces 1. MESA Iteratioal, Metrics That Matter, www.mesa.org, Chadler, Ariz., October 6.. OPC Foudatio, OPC DA 3. Specificatio, Scottsdale, Ariz., March 5. 3. Stephaopoulus, George. Chemical Process Cotrol, Pretice-Hall, N.J., 1984. Author George Buckbee, P.E. is director of product developmet at ExperTue, Ic. (1 James Drive, Suite A, Hartlad, Wisc.; Phoe: 6-369-7711; Fax: 6-369-77; Email: george.buckbee@ expertue.com). George has over years of practical experiece improvig plat performace. He holds a B.S. Ch.E. from Washigto Uiversity (St. Louis) ad a M.S. Ch.E. from the Uiversity of Califoria, Sata Barbara. Tired of promotig your compay with the usual gimmicks? The check out article reprits from Chemical Egieerig! Impressive ad uique, Chemical Egieerig Article reprits are a effective marketig tool for positioig your compay or products. Perfect for tradeshows, media kits ad cofereces.persoalize your reprit by usig a array of optios such as addig your compay logo, advertisemets or cotact iformatio. For more iformatio, please call 8-777-56 or 31-354-11 Fax: 31-39-3847 email: clietservices@accessitel.com 58