Soft sensors — known also as software sensors or neural-network-based inferential calculators — are operators’ and engineers’ virtual eyes and ears. “Software sensors create windows to your process where physical equivalents are unrealistic or even impossible,” explains Jan Versteeg, process control consultant at SABIC-Europe’s Manufacturing Competence Center in Geelen, The Netherlands. The sensors also enable more frequent measurements than possible with hardware, adds Warrington, U.K.-based Paul Turner, Aspen Technology Inc., Cambridge, Mass.
These inferential calculators have broad applicability, says Don Morrison, product marketing manager with Honeywell Process Solutions, Phoenix, Ariz. For example, refining applications range from simple tasks like finding Reid vapor pressure to complex ones such as determining distillation cutpoint or percent light key in heavy product — and everything in between, he notes.
The sensors also predict product quality or combinations, adds Gail Powley, director of strategic initiatives for Matrikon Inc., Edmonton, Alta. The number of inputs may range from a few to hundreds. Sometimes, the input may be just one key parameter, such as distillation tower temperature, she notes.
Outputs typically find one of two principal uses: as open-loop advisory information for operators, or as inputs to model predictive controllers (MPCs) or adaptive controllers for closed-loop control.
“Ten years ago, it was 80% prediction (open loop) and 20% closed loop,” says Ric Snyder, senior product manager for Pavilion Technologies Inc., Austin, Texas, in describing how its end-users deployed their sensors. “Now it’s 80% closed loop and 20% prediction.” He says that as more people became comfortable with the soft sensor, the logical question became: “Why can’t I use this measurement like any other?”
Today, there’s also increasing interest in using the sensors with model-free adaptive controllers and intrinsically safe, nonlinear models. The sensors also promise to play a role in providing multivariable outputs.
Before any soft sensor can be deployed, however, users must address the quality of data to build it. That requires data cleaning — and such preprocessing is a major need, suggests Dave Shook, Matrikon’s chief technology officer. “When building a soft sensor, it’s important that you get all the corrupt data out.”
Most plants have automated data historians that collect and then usually compress data. “[But] compression can be a very dangerous obstacle,” explains Morrison. “You need to know where your data came from and what’s been done to it before you start working with it.”
The compressed data can help to identify the major variables associated with a process. “[However,] since this history is often used for near-real-time diagnostics and troubleshooting, it cannot be artificially filtered or compressed since that can hide dynamic behavior that might be important. We really want and advise customers to give us uncompressed data,” Snyder stresses, adding that no technology exists to unfilter a stream of data.
Regardless, users must specify which data will be used to develop the soft sensor — and those data must be accurate, says Morrison.
“You train the model and then use your test data to see if it accurately replicates the process,” Snyder notes. He cautions that data used to construct the model must never be used to train the model.
In addition, end-users need offline verifications, such as analyzer data or laboratory results, to ensure that the virtual sensor accurately tracks the real process. What’s at stake is the sensor’s credibility. “If the operators ever get to the point where they believe the model is no better than their own best guess, then they’ve lost confidence in the model and probably won’t use it anymore,” Snyder says.
Camcari, Brazil-based polyethylene manufacturer Politeno (Figure 1) put in considerable effort to ensure the credibility of its Pavilion soft sensors. “We had to collect data during a long period (before building the sensors), because we produce 60 grades from the lines,” says Jean-Claude Cailleaux, the facility’s technical manager. The company has two sensors on each of two low-density polyethylene (LDPE) lines for open-loop control, and two on its linear-low/high-density line for closed-loop control. The LDPE sensors predict melt index; the closed-loop soft sensors predict that and polyethylene density.
|Figure 1. Polyethylene plant in Brazil gains over 4,000 metric tons of additional output because of easier grade transitions.
The company’s efforts paid off. “We have reduced the (product) variability by more than 20%,” Cailleaux says. That makes shifting to different grades during production much easier and cuts losses. “We’ve reduced by 40% off-spec production during transitions,” he adds, noting that this translates to an additional 4,000 to 5,000 metric tons of saleable product each year.
The chemical industry historically has relied on the sensors mainly to provide valid, actionable and timely data. “Where soft sensors are most useful … is where the companies have analyzers installed and the cycle time is very long, relative to the frequency at which they want to control the process,” states Morrison.
Today, advances in soft sensor technology are extending measurement capabilities. For example, Worsley Alumina Pty. Ltd. (Figure 2), near Collie, Australia, now uses empirically based Honeywell soft sensors to measure Loss On Ignition (LOI), which is the amount of moisture in the final product from its calciners, says Angelo D’Agostino, senior process control engineer. This is the first use of empirically based soft sensors at the site, he explains, noting all previous soft sensors at Worsley were based on first principles. To date, however, a first principles-model of LOI has not been possible, he says.
|Figure 2. Soft sensor supplements twice daily laboratory analysis of moisture in product coming out of calciners at Australian plant.
Source: Worsley Alumina
SABIC-Europe, at polyethylene plants in The Netherlands (Figure 3) and Germany, relies on Aspen Tech soft sensors for measurements usually not made by online sensors. SABIC monitors density of materials in the reactors and granulator outflow, among other variables. “Density measurement [on-line] is not common in our type of industry, so you have got to get samples and send them to a laboratory. You suffer from time delay as well as low-frequency information. So, a software sensor is here a great solution to better control,” Versteeg notes.
|Figure 3. This polyethylene plant in The Netherlands avoids delays in measurement of density incurred by conventional sampling.|
Using a soft sensor typically reduces laboratory analyses. Politeno needed five fewer analysts once it installed the sensors, Gailleaux says. SABIC-Europe saw the same trend, notes Versteeg. Oktay Karagoz, a process control engineer at SABIC-Polyolefine GmbH, in Gelsenkirchen, Germany, explains that soft sensors reduced the number of laboratory samples he needs by 75%.
In addition, the closeness of soft sensors’ predictions to results from laboratory analyses is boosting end-users’ confidence. Often times, the analytical value is not as good as the inferential value, Karagoz says. Indeed, laboratory analyses are more prone to bias, he adds. At SABIC-Polyolefine, he reckons that 95% of this is from the laboratory itself — e.g., from how the analyses are conducted, how samples are transferred, or from residue in sample ports in the process, when samples are taken shortly after the transition from one grade of product to another. The remaining 5% comes from the gas chromatograph.
Closing the loop
SABIC-Europe’s and Politeno’s use of soft sensors as part of MPCs exemplifies the most significant trend in their use: as the foundation of closed-loop control, either through MPCs (Figure 4) or multivariable controllers (MVCs).
Versteeg says that during transitions from one grade of product to another, when no laboratory data are available, SABIC relies on the MPC. Karagoz adds, “We reduced off-spec during transition… And during steady-state — only one product being produced — we don’t have any off-spec.”
Worsley Alumina pared off-spec product by teaming soft sensors and MVC, says D’Agostino. Worsley used the sensors to fill in the gaps between LOI measurements that were produced only twice daily by their laboratory. “If we can replace some of the analyses with the soft sensor, we can improve the control of the plant,” he says.
Sasol Germany GmbH, Brunsebuettal, also improved control (Figure 5). In 2001, it installed Pavilion soft sensors as an open-loop advisor in its fatty alcohols manufacturing line, to predict a property of an intermediate product stream from the reactors in the Zeigler process, explains Helge Timm, the engineer responsible for the facility’s advanced process control. Then, in 2002, Sasol installed eight more soft sensors in an MPC in the Ziegler process’ main alcohol distillation system. The controller predicts component concentrations in column product streams, Timm explains. The columns’ on-line gas chromatographs give values every 15 to 20 minutes, which he calls too infrequent.
|Figure 5. German facility employs soft sensors for closed-loop model-predictive control of its main alcohol distillation system.
Source: Sasol Germany
Use of the sensors and MPC has led to fewer lab analyses, less energy consumption in the distillation columns and better product quality, says Timm. “Soft sensors give the best control,” he believes.
Clearly, soft sensors are destined for a wider role in conventional closed-loop control. In addition, the inferential calculators are making a mark in emerging control approaches.
For instance, Cybosoft-General Cybernation Group Inc., Rancho Cordova, Calif., has developed soft sensors that are used in its model-free adaptive (MFA) control systems to predict the heating value of mixed gas and steam dryness of boilers, says Dr. George Cheng, the company’s chief technology officer. “Since MFA only needs the setpoint and process variable as the input information to compute its control output, it does not require some of the special information that a model-based approach may require.” So, MFA may need just one soft sensor for the process variable, he explains, while model-based approaches may need extra soft sensors to get the special information.
Use of soft-sensor-based MPCs in more-challenging environments is growing through a nonlinear approximating technology invented by Aspen Tech’s Turner in 2000. The problem encountered in nonlinear models, particularly black-box ones such as neural networks, is that the models may appear to be correct in certain regions but not at other operating points, he says. “Historically, nonlinear models that were accurate were ‘black boxes.’ More powerful nonlinear approximators, such as neural nets, have to be used in a pure advisory mode — outside of the controller — so that any potential unsafe conditions are not seen by the controller.”
Turner stresses that his technology is intrinsically safe and estimates it is in 300 to 500 models running globally, particularly in the bulk polymer industry. Conventional practice, he adds, is to use intrinsically safe models for control and that has meant relying on simple linear models or ones with transforms such as logarithms or quadratics. His model always matches the true direction of the process — e.g., it shows positive temperature gain if that is actually occurring — to prevent potentially unsafe control decisions.
Canada’s McMaster University, Hamilton, Ont., through its Academic Consortium, is working with Honeywell, says Morrison, to develop soft sensors that can handle multivariate analyses to detect general process abnormalities — for example, distillation column flooding or temperature excursions and even erroneous thermocouple readings in reboilers. “Multivariate uses principal-component analysis and could look at the outputs of hundreds of hard sensors and even soft sensors,” he says. What differentiates multivariate soft sensors from univariate ones is that the former looks at process quality, while the latter looks at product quality.
Users are specifying soft sensors more and more for control applications from the outset, Snyder says. “If it’s important enough for them to predict [a value with soft sensors], then it’s important enough for them to control.” He anticipates no quantum leap in soft-sensor technology in the next few years, but expects adoption of soft sensors to provide model-based predictions in a variety of new processes and industries, such as pharmaceuticals.
One overall trend is hard to miss: soft sensors will play an increasing role in automated controllers of whatever type.