Magazine   May/June 1998   pgs. 28-31

Neural Networks 
Monitoring, Control & Optimization

Paul Van Buskirk
Quality Monitoring & Control


Advances in AI modeling techniques, such as artificial neural networks, e-Model, genetic algorithms, VB-Model, etc., provide a robust tool for monitoring, control and optimization of a system. Monitoring is used for data verification, fault detection, forecasting and/or prediction. Total system or subsystem performance can be ranked when combined with statistical methods. When variables are adjustable, the system can be regulated to provide the "Very-Best" results through direct or fuzzy-logic control and optimization methods. This is not for the future; this is the capability that personal computers and artificial intelligence provide today.


Model development is an investment. Recent advances in PC and AI systems provide a significant reduction in this investment. And, there are numerous model applications that will produce a return-of-investment. This article gives an overview of model applications, available now, to improve the profitability, quality, safety or environmental controls of a process. Extensions of these applications can be made to any industry that has data.

This article highlights neural networks as a model development tool, due to its widespread use and success. Other AI modeling methods can be used. The model developer should determine the best technique for a given application.


The advantage of AI modeling techniques, such as artificial neural networks (ANN), are that a considerable amount of time can be saved in model generation when data is available. Development of an equation based (fundamental) model for multivariable non-linear systems can be virtually impossible with certain processes. Furthermore, a fundamental model may require hundreds if not hundreds of thousands of equations. The number of calculations and resulting CPU times can exceed the actual system response in real time. Conversely, ANN models are computationally efficient. The code required typically does not exceed three pages, even for large systems.

ANN model development is data intensive. A database, or spreadsheet, contains the data that the ANN model uses. A column in the spreadsheet represents a single variable. A column can be used as an independent "input" or as a dependent "output" variable. With neural nets, multiple outputs are allowed. How the output variable(s) change with changes in the input variables is what the neural net "learns". A row of data across all used variables is an event the neural net will use to learn the cause-and-effect dependence. The ANN model then "trains" until the error between the predicted and actual value, across selected rows of data, is reduced to a satisfactory level. An introduction into how neural-networks work is provided in reference (1).

ANN model generation does not come without risk. The ANN model developer needs to follow certain data screening guidelines to ensure integrity in the data, please see the General recommendations for the AI model developer. The work involved with data screening can occupy the bulk of the time in ANN (or any other) model development. The ANN model software provider will also supply guidelines for software use, data screening, and model development. Numerous technical publications address issues related to ANN model data preparation, PC AI is an excellent resource for these articles.

General Recommendations for the AI Model Developer.

  • All cause-and-effect variables are measurable and used.
  • Redundant variables are minimized or eliminated.
  • The variable data is uniformly distributed with sufficient points.
  • Data sampling times shows the actual system response.
  • The data is properly screened for outliers and redundancy.
  • True variability exists in the data and all data is valid.
  • Model accuracy requirements do not surpass measurement accuracy and variability.
  • Model prediction will not be extrapolated in any variable dimension.

When these guidelines are followed, an accurate and robust ANN model can be developed that can be used for multiple and strategic purposes. A general overview of these model applications is provided in AI model applications and utilization.

AI Model Application and Utilization.

  • Sensor/data error and fault detection.
  • Prediction and forecasting.
  • Sensitivity or stability analysis.
  • Equipment or sub-system performance monitoring.
  • Total process or system indexing (comparison performance monitoring).
  • Control(s) monitoring.
  • Multivariable non-linear process/system control.
  • Multivariable non-linear process/system optimization.

The applications of a system model are not limited to the items listed. Endless possibilities exist. End use will be as varied as the personnel and industries that develop or utilize the model. The applications listed are available now, through numerous software providers. Implementation will provide a technology advantage, with benefits that improve profit, quality, environmental and safety controls.

As shown above, applications of an ANN model can range from sensor & data verification to forecasting to full process optimization. In most cases the required ANN model(s) for these applications can be developed from the same database. For a given process or system, these applications should be used together.

This approach provides constant monitoring of the sensors, equipment, controls, and total system performance. Deterioration in the performance of any component in a system can then be quickly pinpointed and corrected. As a result, the stability of the process will be improved, providing an increased utilization of the control and optimization applications. Maximum utilization of the control and optimization applications will increase the process profit and quality. Additionally, safety and environmental goals will be maintained and improved.

The following provides a brief discussion of selected items as listed in the AI Model Applications and Utilization:

Utilization of ANN models for sensor validation and fault detection is in common use. A reliable fault detection technique that uses several AI modeling tools is described in reference (2). This application is most effective when all sensors are monitored as part of a preventative maintenance program. The keys to successful process control and optimization implementation are reliable and accurate sensors. 

Sensor Error Meter provides an example of a sensor validation and error detection application.

This application shows a typical sensor monitoring system. The lists on the left are the various sensor names. The box graphs show the sensor errors, in percentage. The top graph shows the current error, the bottom graph shows the average error. Only the top four sensors with the maximum errors are given in this example.













Prediction is primarily used for forecasting. Applications include market analysis, sales projections, product properties, process variable inferred values, etc. Reference (3) is a good starting point for the utilization of ANN models in the process industries. There are also several articles in past issues of PC AI Magazine. The major use of ANN models in the process industries is in the prediction mode, either real-time or for off-line analysis and troubleshooting.

Sensitivity analysis is used to determine the important variables in a process. This provides a fundamental understanding of the process in terms of key variable rankings to control a process result. Numerical derivatives (D Output/ D Input) of the ANN model provides the sensitivity results. Most ANN software providers furnish this feature. Sensitivity analysis is classically the first step towards implementing supervisory controls into a process or system. Numerical derivatives of a process provide what is termed the gain array, which is the starting point for multivariable control and optimization.

Equipment performance monitoring can be applied through the use of several approaches. A direct approach is to develop an AI model to determine the performance index of the equipment. The calculated index from the AI model is then monitored to determine the performance decay rate. From the decay rate preventive maintenance requirements are determined using "rule-based" AI technologies.  The Equipment Performance Meter shows a typical screen. This type of approach can be used for the majority or process equipment designs and subsystems.

Equipment Performance Meter

This graphic shows a typical equipment performance meter. The example provided is for a compressor monitoring system. The index monitored is the compressor efficiency ratio, please see graph. The graph provides a run chart of compressor operations. Efficiency ratio decay is used to predict maintenance requirements. Other data gives average and current operating conditions.














Total process indexing provides a single auditing measurement to monitor the performance of a process or system. This method combines the sensitivity of a process with statistical process control (SPC). The auditing measure, "Process Index", is calculated by the following equation (4):

Where xi = data point evaluated at ith sampling of input variable x.

Fi = ANN model prediction at ith sampling point.

= Target input variable setting.

s s = Standard deviation.

= 2.71828182846.

The exponential term is the normalized Gaussian distribution function. The Process Index weights the sensitivities of a process with this distribution function. Input variables with high sensitivities will have a larger impact on the Process Index than those with low sensitivities (for the same departure from the target values).

The Process Index provides focus for the variables that most affect stability or quality in operations. The index value is scaled from 100% (ideal) to 0% (unacceptable). The index should be configured as a run chart to monitor total process performance. An example is provided below.

Process Index for Polymer Manufacture

This Plant Index example shows the results of an analysis for polymer manufacture. The top graph provides the ANN model's prediction versus the measured product melt index (MI). The MI is a polymer property used for sales specification. The highlighted list gives the ANN model input variables. Prod_MI is the modeled, or output, variable. The four top factors that contribute to a Plant Index reading of 64.65% are listed on an instantaneous and average basis. The Plant Index individual values are also listed. In this example Recycle_Flow contributes 16.01% from the product being produced off targeted values, i.e. a Plant Index of 100% is ideal.














Control monitoring provides an application technology that gives the best pairings of controls in a process. With a process that can be regulated, certain variables can be independently manipulated, termed control inputs. These control inputs are adjusted to control certain system requirements and/or specifications, termed control outputs. Control monitoring provides the best pairing of manipulated-to-controlled variables.

The technology is based on an ANN model to generate the gain array (from process sensitivities). Minimum interaction of manipulated-to-controlled variables is the goal (5). The maximum performance occurs when interactions are eliminated. The performance indicator is named the "Stability Index". A value of one is ideal. Negative values indicate an uncontrollable control scheme. High positive values indicate a system that is marginal.  Multivariable control analyses shows an industrial application of this technique.

Multivariable Control Analyses

The control monitor application, as shown below, provides the best pairings for process regulation. This example is an industrial distillation column. Distillation columns are used for separating a single stream, with two or more components, into two product streams of higher purity and value.

In this example, the control inputs are Dist_Flow, Btm_Flow and Reflux_Flow. These labels are for the two product streams termed distillate and bottom, and for a stream that is returned (refluxed) back into the column. The control output variables are Btm_Conc, Dist_Conc and Column_DP. These terms are the bottom concentration, distillate concentration, and column differential pressure, respectively. The product streams worth is based on their concentrations, and high column differential pressure can lead to a process upset. Control of these variables is required for successful operations. The goal is to control this process near maximum, with variable, rates.

Control (quality) monitor results show the best pairing combinations (see Input to Output Pairings). The Stability Index value is in a range that indicates these pairings will be stable. Therefore, with this multivariable control configuration, interactions will be minimal due to set point (target value) changes and external disturbances.














Multivariable non-linear process control and optimization methods directly follow from the methods presented. Once a neural net model is developed, the equations can be imported into several optimizers, such as Excel’s solver. Scientist , by MicroMath provides a wide range of optimization methods that directly interfaces with several spreadsheet programs.

However, for "real-time" control and optimization applications, a time dependent-dynamic ANN model is needed. Reference 5 gives an overview for several "adaptive" or other state-of-the-science methods. Additionally, multi-variable linear and non-linear process control and optimization methods are currently available from several software developers. Again, PC AI is an excellent resource in this field.


A system model provides a base for numerous applications. These applications can be used to improve the stability and control of a process or system. This will improve profit, quality, safety and environmental objectives. Applications using available technologies include sensor, equipment, control and total process monitoring. This can lead to successful implementation of supervisory controls and optimization applications to maintain maximum performance.

PC & AI systems and technologies allow for rapid and effective model generation and application. As demonstrated, these technologies are currently available and can be implemented now.


Bhagat, P., "An Introduction to Neural Nets.", Chem. Eng. Progress, p 55 (Aug 1990)

Smith, S., "SDI’s e: Real Time Prediction in the Chemical Industry", PC AI, p 18 (Jan/Feb 1998)

Chitra, S., "Use Neural Networks for Problem Solving", Chem. Eng. Progress, p 45 (April 1993)

Van Buskirk, J., "Modeling of a High-Density Polyethylene (HDPE) Process", Texas A&M University, Masters Thesis (Dec 1996)

Ogunnaike, B. and W. H. Ray, Process Dynamics, Modeling, and Control , Oxford University Press, 1994