The chemical and petrochemical industries traditionally have been highly receptive to new control and supervision technologies. These industries have invested more money and time in research and innovation than any other industrial sectors.
Nevertheless, objective benchmarking criteria ," using the more innovative manufacturing sectors (automobiles, electronics, etc.) as a basis ," show that the efficiency of chemical plants is still far lower than its potential. Improvements could be achieved through the use of advanced computer technology.
Throughout the history of industrial control, automation gradually has progressed toward higher levels of autonomy, the degree to which operations can be handled without human operators. In the process industries, model-based control has led to enormous progress in autonomy during the past 30 years.
Process control began with simple configurations made up of a group of single-loop controllers, with little or no reciprocal coordination. Today, multivariable controllers optimize larger and larger production units. One "package" oversees functions that once could have been implemented only through the use of several different strategies (single-loop control, anti-windup, uncoupling to remove multivariable interactions, constraints management, delay time compensation, feed-forward management, economic optimization). This dramatic function integration, permitted by improvements in both the mathematical algorithms and the processing and memory capability of processors, means that human operators can focus on high-level actions and decisions.
In the past, to control a distillation column using numerous single-loop regulators, operators had to set and monitor each single set-point value for flowrate, pressure and other variables that were only implicitly correlated to the end product's quality. A modern model-based architecture manages these execution details automatically and allows the operator to run the plant by generating commands directly related to the quantity and quality of the end product. One single operator can control a larger portion of the plant. It is possible to measure the level of autonomy of a control strategy by means of a special metric defined by the "number of loops to be supervised per operator." The higher this parameter, the higher the autonomy of the system implemented.
Control systems autonomy has had an enormous impact in economic, environmental and human terms. In 1980, the U.S. refinery industry produced 5.3 billion barrels and employed 93,000 people. By 1998, production in this sector rose by 17 percent, while the number of employees dropped by 35 percent to approximately 60,000.
The first on-line computer-based control system (103 inputs and 14 outputs) was put into service on a catalytic polymerization unit at a Texaco plant in Port Arthur, Texas, in 1959. In the early 1970s, multivariable model-based control was developed in parallel and independently in France and at the Shell research laboratories to optimize petrochemical and refining units.
A census of industrial multivariable control applications by Qin and Badgwell in 1997 shows that 89 percent of these applications occurred in production units within the chemical sector, including refining, petrochemicals, specialty chemicals and pharmaceuticals. A second Qin and Badgwell survey on the introduction of even more innovative techniques in this field placed more than 82 percent of nonlinear control applications in the chemical sector.
In 2000, the chemical industry invested just under $220 million in process simulation and optimization systems ," 65.2 percent of the total investment ($336 million) by all industries, including power, hydrocarbon extraction and distribution, food and beverage, pulp and paper, steel and iron and more. Yet, only 37 percent of total chemical-industry investments were in industrial automation.
Despite the huge investments it has made in information technology, the chemical industry has been unable so far to achieve the performance of many other manufacturing sectors. Why? Experts have suggested a number of reasons, including:
Knowledge deficit. Contrary to what occurs in discrete event systems, detailed knowledge of the production process often is not available in the chemical industry. The lack of full access to physical-chemical details gives rise to a relatively high degree of unpredictability and, therefore, of unreliability.
Extreme technological diversification. It is difficult or impossible to standardize or fully duplicate procedures, technologies and operating practices.
Riskiness of production. Labor-intensive, high-risk chemical processes with a high impact on the environment encourage a natural ," and understandable ,"
conservatism inside plants. This conservatism can result in resistance to change.
Limited production flexibility. The average plant is large, complex and costly, and any changes to the setup and/or product are very complicated. Quick responses to economic trends or changing market demands are difficult.