On the Programming of Industrial Computers

This report which is part of the deliverable IP of the Esprit project VHS Veri cation of Hybrid Systems analyzes some software engineering aspects of industrial computers such as PLCs programmable logic controllers and DCS distributed control systems The report starts with a comparison between the development of software engineering for general purpose computers and the programming of control computers Then it critically surveys the ve programming languages de ned by the IEC standard which is intended to unify PLC programming languages Finally several potential contributions of the consortium toward improving the state of the art in this domain are suggested Introduction Industrial vs General purpose Computers Programming methodologies for the general purpose computer have undergone tremendous improvements since the s In few decades programming moved from machine code and assembly languages via the rst high level languages structured programming data types cross compilers and debuggers up to object libraries graphical programming environments and inter application interfaces The outcome of all these developments and many others not mentioned here is that programming can be done at a more abstract and problem oriented fashion letting the computer itself do the tedious tasks of handling the technical details of the speci c hardware platform on which the program will eventually run Consequently the cost e ectiveness and quality of software production has increased beyond what could be imagined in the early days of computers So far the development of programming methodologies for special purpose industrial com puters has been signi cantly slower Before we analyze this problem its origins and hope fully some of its solutions we have to characterize even roughly the type of systems we are talking about First and foremost we are concerned with systems whose major role is to inter act via sensors and actuators with a dynamic physical environment in other words computers that control Of course ordinary computers are connected as well to peripherals such as a mouse a keyboard or a communication port but this interaction is not their raison d etre The functionality of control computers is de ned in terms of the performance of the physical processes with which they interact In order to achieve this performance the control system monitors input signals which deliver information from the process makes some calculations While the scope of this report is the class of systems controlled by PLCs and DCSs some of the initial considerations are common to all computer that control sometimes referred to as embedded reactive or real time systems and produces output signals which are then ampli ed and transduced into actions that in u ence the process The calculations done by the controller are usually simple compared to arbitrary algorithms which can be performed by general purpose computers but they must respect some timing constraints due to the interaction with a dynamic environment which does not wait for the controller to conclude its computations One can distinguish between two types of signals continuous and discrete The former represent quantities such as temperature velocity or water level while the latter can stand for on o states of devices such as valves and furnaces for threshold conditions on continuous quantities or for higher level supervisory signals In the pre computer days calculations over continuous signals where done by analog means rst mechanically and later electrically by analog computers This is the origin of the block diagram paradigm where input signal move through arithmetical components and integrators to produce output signals Similarly the discrete logical input output functions were computed initially by electro magnetic and pneumatic relays and later by hardwired digital electronics which can be viewed as a discrete version of block diagrams The rapid progress in VLSI technology and the arrival of cheap micro processors a ected both types of control systems For continuous control it turned out to be more cost e ective to replace analog noisy devices by digital calculations The price of going digital is the replace ment of real numbers represented by magnitudes of physical quantities by their binary en coded oating point approximations and of continuous control by sampled piecewise constant control The nature of continuous processes sampling theorems combined with the speed and accuracy of digital computers guarantee the dominance of digital control Discretized versions of continuous blocks such as integrators and PID controllers are widely used On the discrete control side the replacement of the hardwired and later programmable logic controller by a general purpose microprocessor was even more natural With a microprocessor inside one would expect that control computers will converge toward general purpose com puters and bene t from the progress in software engineering but some technical conceptual and sociological factors are slowing down this convergence To analyze the situation let us see what is still special about computers that control The major feature of control computers is the need to interact with many I O signals com ing from various types of devices From the computer hardware point of view this means that the processor board is augmented with specialized I O boards whose function is to connect physical signals with the computer memory using A D and D A conversion ampli cation of output signals in order to drive actuators and communication with the processor In addi tion these computers are supposed sometimes to work in hard physical conditions and be more solid and reliable Hence the processor and its software constituted a secondary concern relative to measure ment technology and other electrical and mechanical engineering considerations The intimate connection with the physical world explains the lag in development of more high level pro gramming culture in this domain In fact the situation is similar to that of operating systems programming two decades ago where systems programming was dominated by assembly languages and low level primitives to handle I O devices and improve performance The diversity of plants to be controlled such as airplanes and missiles re neries CNC machines production lines at various scales railway systems washing machines alarm sys Simple in the technical sense that all instances of the computation can be performed within a priori bounded time and space These computations can be very complicated in other senses tems and what not made it hard to reach a general abstract concept encompassing the common features of all these phenomena As a result real time programming also known as embedded systems computer enabled control etc is one of the most eclectic and con fused parts of computer science In some safety critical domains such as avionics or nuclear plant monitoring there was a concentration of a critical mass of software engineers working on the same application using the same hardware and software platform However in many industrial domains software developers work in relative isolation they are not the domi nant sub community in the enterprise and their productivity is not the determining factor in hardware selection decisions Consequently they are often locked in the speci c programming environment supplied by hardware vendors and need to stay with the same vendor in order to reutilize already developed software This situation was very convenient for the major hardware vendors In contrast the general purpose market converged mostly into the open architectures of the personal computer or workstation where hardware vendors compete with each other o ering a variety of choices in motherboards processors I O cards modems monitors etc This open plug and play architecture encourages competition among vendors resulting in price reduction and can be viewed as one of the driving forces behind the proliferation of computers in contemporary society The quest for a similar standard interface is now taking place in the consumer electronics market IEEE and will probably change the landscape of this domain as well As for software albeit the dominance of Microsoft on the Intel architecture users have a choice in almost any application domain including operating systems and there are numerous programming environments including several competing compilers for every useful programming language Inspired by this state of a airs users and developers of industrial control systems started to push toward an open hardware and software architecture for industrial computers based on a PC like computer with a modern programming environment The advantage of such PC like systems is their easy integration with other software used by the enterprise This connection is achieved either by interfacing the PLC operating system with a commonly used operating system or by using the concept of a Soft PLC which is deliever the functionality of a PLC as a software task under an ordinary operating system At the hardware level compatibility is to be achieved by the use of one or more out of several standard buses and communication protocols already being used in the industry some eld bus standard already exist for DCS In order to create a common base for stadards terminology and software the IEC International Electro technical Commission formed several technical committees resulting in the so called IEC standard published in Later an independent association called PLCopen was created to promote the usage and supply of products in conformance with this standard Compare again with the much simpler domain of business data processing in which it took many years to develop the unifying concept of the data base management system separating the logical from the physical in information storage This is part of a more general standard supposed to de ne common terminology in the chemical process control world but we will concentrate on the software