mation systems and the need for access to system data. The paper then describes the development of DEC @aGlance software and the choi ce and use of Appl ication Control Architecture (ACA) Services to solve the problem of integrating independently developed applications i n the manufactu ring space . 2
Background
I n large manufacturing facil ities, the production process is contro l led through the use of advanced automation systems. These systems may track thou sands of temperatures, flows, pressures, and levels and can drive hu ndreds of pumps, valves, and other actuators. To implement control strategies, such systems may compute large numbers of complex , dynamic control a lgorithms. Usu a l ly, additional sys tems measure various physical properties of the product, such as color, weight, viscosi ty. thick ness, and moisture content. Supervisory control systems
DEC @aGlance-Integration ofDesktop Tools and Manufacturing Process Information Systems
often coordinate parts of a complex process, as wel l as implement h igher-level control and produc tion strategies and keep h istorical records of key process variables.
The control of a large plant is usual ly imple mented through strategies that al low the control problem to be d ivided i nto smal ler parts, as i l lus trated in Figure 1 . Each piece of the system is responsible for the control of a subsystem (e .g., steam generation and d istribution, or cool ing flu ids), a part of the process (e .g. , premix i ng, material storage, or reaction), or an area of the plant (e .g., packaging l ine, product stream, or finished goods management). Within each subsystem, there is typi cally a h ierarchy of control. The lowest- level com ponents control activities that require responses within less than a second to as much as one minute (direct control). The next level of systems control activities that require responses within Jess than a few m inutes (d istributed control). Above this level of response are systems that control activities that may not change for long periods or t hat i mple ment control algorithms that involve measurements from more t han one lower-level system (super visory control). At the plant level, additional
PROCESS HISTORIAN HISTORICAL
�
PROCESS DATABASE REAL·TIME PROCESS DATABASE DISTRIBUTED CONTROL SYSTEM ADVANCED CONTROL LOOP CONTROLLER LOOP CONTROLLERPUMP GAUGE VALVE
control systems may exist to i mplement control algorithms t hat reflect changes in the markets for products, market opportunities, and fluctuations in raw material avai labi lity and composition, a long with the information about the process that is sup pl ied by the lower- level systems (h igh-level con trol). Scattered among these levels may be various additional systems that schedu le p reventive main tenance, identify equipment fai lures, and advise on process improvements-all based on i nformation abou t process from the other systems in the plant. D istributed control systems include an operator console that consists of multicolor d isplays, push buttons, warning l ights and buzzers, a touch screen or trackbal l , and i ndustrial ized keyboards with as many as a 36 special fu nction keys. The d ispl ays al low an operator to oversee al l parts of the process for which the operator is responsible. Typ ical dis plays show recent trends of key variables and mimic diagrams showing the current state of the manufac turing equipment (e.g . , valve positions and tank lev els) and of the material flowing through the process. The keyboard and other i nput devices allow the operator to select d isplays, request reports, and mod ify control settings. Response to
ARTI FICIAL INTELLIGENCE SYSTEMS H IGH·LEVEL CONTROL SUPERVISORY CONTROL DISTRIBUTED DISTRIBUTED CONTROL CONTROL SYSTEM LOOP CONTROLLER DIRECT CONTROL SENSORS AND ACTUATORS
Figure 1 Typiectl Levels of Control in a Process Plant
Application Control
problem or alarm concl i tions and mod ification of the p rocess to change the product are effected through the console.
Process operators are responsible for maintain ing the routine operation of a plant. Operators use the control system to change process parameters in order to produce d ifferent m ixes or variants of the product, or to respond to an equ ipment failure by rerouting material around nonoperational p rocess equipment.
To perform their functions, manufacturing plant production and engineering support personnel (e.g. , control engineers, process engineers, produc tion supervisors, production planners, mainte nance supervisors, and manufacturing engineers) also need access to information in the control and supervisory systems. These professionals regu larly access inforn�ation contained in multiple manufac turing systems and have an occasional interest in particu lar measurements or parameters within other parts of t he process. The functions of t hese manufacturing plant personnel incl ude
• Complex problem analysis and solution.
Locating sources of product or process variation involves analyzing i nformation from different parts of the process that may be under the con trol of d ifferent automation systems. Comparing the flow that exits one parr of the process with the flow that then enters the subsequent part, for example, could disclose a fau l ty flow meter, a previously unknown temperature con trol problem, or a leak.
• Product i mprovement. I mproving prod uct qual ity and consistency involves investigating how the produ ct is affected by existing variations in the production process. For example, investi gation m ay involve the study of a process vari able that cannot be measured directly but can be calculated from the values of other process vari ables. Examining sets of variables over time and exploring possible relationships m ay resul t in discovering combinations of process variables that yield u nexpected effects on prod uct attribu tes.
• Process i mp rovement. Improvements i n process yield and process rel iabil ity and reduction of waste and hazardous by-products may involve the study of h istorical data va lues from the pro cess. Studying measurements obtained from mul tiple control systems may also result in pro cess i mp rovements.
102
• Resource optimization. Usual ly, process plants are capable of produci ng different grades of prod uct, as we l l as mixtures of end products. An oil refine ry, for example, produces various grades of fuel oil and also home heating and lubricating oils, aJJ from a si ngle process. While the operators adjust the equipment to control the product mix, a process planner or produc tion manager determines the best production schedule based on customer orders and the effi cient use of the process equ ipment.
Process information is available to operators and engineers who are trained to work w ith the various control and management systems in the plant. Using proprietary tools for each system a llows reports to be generated and specific types of analyses to be performed on the data contained within each of these systems. However, extracting the data from these systems to an engineer's desk top for analysis by generic tools, such as spread sheets and statistical analysis packages, is difficu l t or even impossible. Lack o f console- a m i tool specific tra ining is another obstacle to accessing process information.