Predicting Product Quality with Soft Sensors: Application, Opportunities and Challenges Overview of Most Popular Applications
Chemical plants are usually highly instrumented and have a large number of sensors that collect measured data for process control and monitoring. About two decades ago researchers began using the large amount of data to build predictive models, and these models were called soft sensors. The term soft sensor is a combination of the words ‘software’, because models are developed in computer programs, and ‘sensors’, because these models are providing similar information as hardware sensors. From then, soft sensors have found their wide application in on-line monitoring and optimization of chemical plants.
Soft sensors are most often data-driven models based on data measured within the processing plants, providing real-time information necessary for effective product quality control.
The span of tasks performed by soft sensors is quite broad but the most common use is the prediction of process variables that can only be known either at low sampling rates or through off-line analysis. These variables are usually very important for overall process operation because they are usually related to the product quality and it is necessary to deliver additional information about these variables at higher sampling rate and lower financial impact.
Another field of application of soft-sensors is of process monitoring and process fault detection by finding the state of the process and identification of the deviation source.
For successful monitoring and control of chemical plants, there are important quality variables that are difficult to measure on-line, due to limitations such as cost, reliability, and long dead time. These measurement limitations may cause important problems such as product and/or quality loss, energy loss, toxic byproduct generation, and safety problem. These are all challenges that can be succesfully bridged with the application of soft sensors.
Soft sensors are mostly used for continuous processes and for process units that include distillation columns, fractionators and reactors because of the complexity, importance and huge energy consumption of these process sections.
Many successful examples of soft sensor application can, therefore, be found in oil and gas and petrochemical industry, specially related to process optimization activities. Improving monitoring and product quality control with continuous, on-line and accurate property prediction is amongst the key steps for optimization of whole units.
Soft sensors for distillation and fractionation units
Soft sensors for Crude Distillation Unit
When referring to oil and gas industry, Crude Oil Distillation Unit with its distillation tower is the heart of any crude refinery because it is the process in charge of the separation of petroleum cuts. These cuts are later being processed in other operation units in order to refine and blend the gasoline, gas oil and other commercial products.
Although there are further technological processes that deal with refining these initial properties, the specification requirements for some of them are very strict even at this initial stage of the process. Moreover, this initial separation affects the efficiency of the whole refinery process due to the fact that the yield obtained in these cuts contributes significantly to the overall refinery profit.
Prediction of the properties of the crude oil distillation side streams such as ‘95% ASTM’ distillation curve specification based on different mathematical data mining methods has been around for decades. However, there are still many problems with the existing estimators that require a development of new adaptable techniques for an on-line monitoring and prediction of the distillation process product qualities. The nature of non-linear characteristics of the distillation, the variety of properties to measure and control together with all the difficulties to identify, control or compensate the dynamic process behavior and the errors from instrumentation for an online model prediction are only some of the problems that a prediction technique should deal with in order to be useful for a practical application.
However, there are many successful application cases that refinery units use real plant data to calibrate models. They can be used to predict quality properties of the gas oil, naphtha, kerosene and other products of a crude oil distillation tower. Some of these are distillation end points and cold properties, such as freeze and cloud point and if developed and used properly can bring significant economic benefits to a refiner.
One of the most important properties is ‘95% ASTM’ distillation curve specification, which limits the amount of product that can be extracted in a side stream. The economic goal of operating a distillation tower is based on obtaining the maximum amount of product with the highest quality within the specification.
Soft sensor in a monoethylene glycol plant
Mono-ethylene glycol (MEG) has emerged as a very important petrochemical product as its demand and price have been considerably rising all over the world. It is extensively used as the main feed for the polyester fiber and polyethylene tere-phthalate plastics production. UV is one of the most important quality parameters of MEG and it indirectly represents the impurities level such as aldehyde, nitrogenous compound, and iron in the MEG product.
In Glycol plant the MEG is drawn off from MEG distillation column as a product, its UV transmittance is affected by many things such as impurity formation in non-removal and accumulation of aldehyde in the system etc. Since these UV deteriorating impurities are in ppb ranges, they are very difficult to detect during the MEG production process and they have hardly any effect on process parameters. That is why it is very difficult for any phenomenological model for UV prediction to succeed in an industrial scenario.
Normally, online UV analyzers are not available to monitor the product MEG UV analysis in an ethylene glycol plant. Off-line methods for MEG quality control is a common practice among the manufacturers, where a sample is withdrawn from the process and product stream for laboratory analysis several times a day and analyzed by time consuming laboratory analysis.
In the event of a process malfunction or operating under a suboptimal condition, the plant will continue to produce an off-spec product until lab results become available. For a big world class capacity plant this represents a huge amount of offspec production results and enormous financial losses.
This necessitates the online UV sensors or analyzers which can give UV continuously on a real time basis. Accurate, reliable and robust UV soft sensors can be a viable alternative in this scenario. The industry needs this mathematical model to predict MEG UV on a real time basis so that process parameters can be adjusted before the product goes off specification.
Soft sensors for reactor monitoring
Reactors are representing the hearts of many units in chemical industry and are the key spots for optimization. Reliable online quality prediction of reactors chemical processes often encounters different challenges, including process nonlinearity, complex interactions between variables, and strict product quality requirements.
Soft sensor for predicting sulfur content in diesel product
When it comes to refining units, hydrotreatment processes are very important in achieving the requirements of low-sulfur fuels. Hydrodesulphurization of gas oil fractions is commonly accomplished in the trickle-bed reactors where there are three phases: gas (mostly hydrogen), liquid (gas oil) and solid catalyst particles.
Soft sensors have proven their predictability of sulfur content in product diesel and can work very efficiently in parallel to a hardware analyzer as a backup while their measurement becomes impossible due to instrument failure, maintenance or repair, or can be applied as the independent measurement.
The appropriate mathematical tool is chosen with regard to the nonlinearity of hydrotreatment reactor and taking into account significant time delay.
Expected results for the sulfur in diesel content are shown in the picture below.
Due to large-scale application of catalysts in the petroleum and petrochemical processes, the main problem concerning these processes is the deactivation of the catalyst. As the catalyst activity is related to the availability of catalyst active area for the reactants, deactivation refers to reducing the activated area of the catalyst or blocking the path of the moving reactants and the products. Mathematical modeling of catalyst deactivation is the main concern of these processes during operation. One of these processes with high potential of deactivation is the disproportionation (DP) of toluene and/or transalkilation (TA) of C9 aromatics and toluene to C8 aromatics over zeolites in fixed-bed reactors. DP and TA involve solid-acid catalytic processes which occur over medium-pore zeolites.
A soft sensor was designed by Hamed Gharehbaghi and Jafar Sadeghi ("A Novel Approach for Prediction of Industrial Catalyst Deactivation Using Soft Sensor Modeling") for identification of the variations in the catalyst activity using process data of an industrial reactor.
The catalyst activity has been estimated non-parametrically as a function of time, olefin concentration, and hydrogen concentration. Other parameters are considered to be constant.
The model prediction is in good agreement with the process data. The present work can be considered as an advance towards the identification, modeling, and fault detection of a complex industrial process.
Challenges in application
Soft sensors are widely used to realize efficient operations in chemical processes because some governing variables, such as product quality, cannot be measured directly through hardware in real time. One of the design problems of soft sensors is the degradation of their prediction accuracy over time.
Models to inferred specifications from available plant measurements, such as temperature, flow rates, pressures, have errors from both the measurements errors and from the inadequacy of the model structure. The plant has a dynamic, non-stationary behavior, it evolves, but the model is fixed.
To reduce degradation, a range of adaptive models has to be developed. In addition, sometimes also a multivariate statistical process control methods must be developed to help select the appropriate adaptive model for each process state. With those methods, higher predictive accuracy than from traditional models can be achieved. This approach may also be used to reduce the maintenance cost of soft sensors.