The experienced team at Aurel can analyze your plant's operational needs and your concerns, create a verified, first-principles, dynamic simulation model of the base process, construct a DDR model to run on top of the dynamic model, integrate the system with your data historian, add optimization, product tracking or advanced controller solutions (as required) and deploy the complete online system, with (optional) monitoring and maintenance.

DDR is a core technology for our online application developments. It makes it possible to fill in missing data and reconcile inaccurate measurements, resulting in reliable and complete data. It provides accurate process information that can then be used to initialize a first-principles simulation model that can be used to monitor and predict process trends and behavior.

Plant process information systems provide access to live and historical data that can be used for many valuable purposes, such as:

• Enhanced energy usage monitoring
• Emissions compliance
• Product tracking
• Process optimization
• Advanced process control
• Improved plant operation and decision making

However, the quality of this data can be negated by the limitations of the plant's measurement instrument technologies, instrument maintenance practices and completeness of coverage. What this means is that plants often lack the complete and accurately measured data that would be required to generate actionable plant information.

Aurel's unique Dynamic Data Reconciliation (DDR) ensures that the online applications that we develop for you will yield complete, accurate and actionable data for better process control, optimization and accounting.

Only Aurel has a mechanism for reconciling data with a simulation model that includes process dynamics. This provides a realistic snapshot of the process in close to real time that can be used to assist operators by predicting process outcomes while accurately modeling the current state of the process.

Aurel's approach to doing data reconciliation is unique. Other companies use simplistic or incomplete steady-state process models for the basis of their optimization products. Our dynamic process models are complete, robust and deliver the accurate, time-based data that is required for informed decision-making.

Aurel's DDR is included as a standard component with every license of CADSIM Plus at no extra charge.

A set of reconciled data based on carefully selected measured data represents a possible state of the process. In practice, it can be assumed that the reconciled data can constitute a reliable approximation of a real process and can therefore be trusted to calibrate a simulation, if there are sufficient numbers of measured data points of good quality, set within ranges that are reasonable for the process.

DDR generates a coherent set of calculated data which is as close as possible to the measured data in the least squares sense. The computational technique consists of minimizing the differences between calculated and measured values under the process equations constraints and is weighted by the standard deviations of the measurement errors.

In order to use Aurel's DDR, you must first build a completely specified CADSIM Plus dynamic simulation model. This model should be tuned and a start up file should be saved, for starting the model from a known set of conditions. Live starting values can be obtained directly from a process historian or a DCS system.

Once the model is complete, you use measurement type specifications to replace selected free variables in the model at runtime. They can be linked internally to other process variables within the simulation model, or more likely, they will be linked via OPC to external variables from a process historian or other system.

Each measurement specification includes a number of options and settings that are used to determine how CADSIM Plus treats the data. These settings include:

• Target value
• Weight
• Change weight
• Typical measurement range
• Specification range
• Time delay (optional)
• External link information (optional)
• OPC quality check (optional)
• Filtering (optional)

You can add as many measurement specifications to the model as you like. It is preferable to over-specify the model when using DDR measurements. This assists the solver to converge to a solution. In practice, there should be a minimum of one measurement specification for each known value or measurement instrument in the real process.

Measurements can use implied relationships that are supplied by the user. For example, if a hot stream and cold stream are mixed, and their temperatures are known, then the downstream temperature can tell you something about the actual flows of the mixing streams, even if no flow meters are installed.

When the model is run, CADSIM Plus begins a number of trials using a simplex (or other optional) method to optimize all of the variables. Each time it completes one trial, it records the error and then rewinds the simulation back in time, and then tries a different starting point. The selected optimization algorithm continues to run trials until the resulting error is minimized and the specified time period has expired. The DDR 'try and rewind' method assumes that data from the 'live' process is obtained in real time. This gives CADSIM Plus the opportunity to perform as many trials as possible before the next update of live data from the DCS system. Usually, CADSIM Plus will be able to perform hundreds of trials and minimize the error to a reasonable level before the next iteration. Trial periods of up to five minutes are typical.

Once the incoming data is successfully reconciled at the end of a trial and the model is balanced, the reconciled values can be put back to a separate set of tags within the process historian system, via CADSIM Plus's OPC or another connection method, for operator review.

A CADSIM Plus model with DDR can be automated via a customized, external launching application which can be set up to run unattended on a dedicated workstation. Ask us about how we can put Dynamic Data Reconciliation to work for you in a custom online application.

Inaccurate process measurements can frustrate attempts at making sound plant operation decisions. Sometimes it is obvious that a measurement is not correct. For example, a measured negative flow, negative temperature or negative pressure of steam flowing to a heat exchanger is likely to be erroneous. However, experience alone does not always allow you to determine whether or not a measurement is correct.

There are many things that can contribute to measurement error. The accuracy of a measurement may depend on the type of sensor and how difficult it is to infer what the measured value should be from any given measurement method:

• Scaling errors (the measurement always reads some fractional amount of the true value)
• Offsets (the measurement is shifted high or low but the change in measurement is reasonably accurate)
• Stickiness (if mechanical, the measuring device changes in steps rather than moving smoothly)
• Intermittent measurement noise (perhaps due to transmission interference)

Often with flow measurements, which vary with the square root of a pressure drop, there are errors at or around zero (when the flow should be zero, it reads some negative value or has some positive residual). In addition, errors are often categorized as being either random or gross. Many permutations of the problems mentioned above are possible, and any of their contributions can also drift with time as sensors age and wear, or as unknown process variables, such as ambient or process temperatures, influence the measurement.

Data reconciliation can help rationalize the measurements of a process in order to form a more consistent and accurate basis upon which to base calculations and decisions. It does this by putting a best fit of the results of a trusted, first-principles process model, onto the change in historical data measurements.

Data reconciliation is the process of taking a set of potentially inaccurate measured values, and then determining a complete set of reconciled values by fitting the plant data to a trusted, first-principles simulation model of the process.

Dynamic Data Reconciliation (DDR) fits a dynamic simulation model (the results of basic conservation law calculations) to the measurement data. These basic conservation law calculations are often referred to as “first-principles” models and involve material and energy balances that include dynamic or transient behavior.

Using first-principles to qualify the data has many advantages over any method that lacks this level of rigor. An example will illustrate why this is true.

Consider a hot water tank with a 10 minute retention time that suddenly gets a step change in one of the input flows that is 10 °C hotter than the other inlets to the tank. Measurements taken a minute later might indicate that the hot inlet has increased in flow, but the outlet temperature of the tank may have changed only a small amount, if at all.

If only steady-state mass and energy balances are used, the outlet temperature (which we will assume to be a reliable measurement) would tend to indicate that the inlet change in flow might not be correct, because tanks in steady-state simulation often have no volume to account for their current accumulated contents. That would be a wrong conclusion.

However, if a dynamic simulation is used, then the time constant of the process would be included in the model and it therefore would be expected that the outlet temperature of the tank would only have increased a nominal amount over the given time period.

Measurements of individual variables that are in dynamic transition can cause completely wrong perceptions when doing data reconciliation, unless dynamic process response is also taken into account. And most processes seldom reach a true steady state, so reconciliation that assumes only steady-state operation can be problematic. This is especially important for real-time online process tracking.

Therefore, fitting the results of a robust dynamic process simulation to the measured plant data offers a better way of getting the most out of data reconciliation because it takes into account the first-principles of mass and energy balance, process dynamics and the process topology.

In addition, the use of a simulator allows filling in all the missing flows, temperatures and pressures for intermediate streams, for which no measurements are available. Any number of those verified values can then be recorded in the process data historian when they are generated by a trusted mass and energy balanced model.