SurveyFit (User Guide)
Overview
Met Dynamics SurveyFit is a parameter estimation tool for fitting steady-state process simulation models to plant survey data. The software integrates directly with SysCAD flowsheet simulations and automatically adjusts model parameters so that simulated results reproduce measured plant data.
This page provides the User Guide for SurveyFit, including installation, workflow guidance, and interpretation of calibration results. For detailed descriptions of commands, fields, dialogs, and options, see Reference. For worked examples, see Examples.
SurveyFit is primarily intended for calibration of mineral processing circuits such as crushing, grinding, flotation, and other beneficiation processes. When used together with the Met Dynamics Models library, parameter estimation can be applied across complete mineral processing flowsheets including both comminution and downstream beneficiation processes. This allows unit operations across different processing stages to be calibrated simultaneously within a single consistent process model.
Because the flowsheet structure enforces mass conservation and model equations, the calibration process naturally respects the physical constraints of the process. The resulting parameter set therefore produces a model that is both consistent with measured plant data and internally consistent within the flowsheet model.
Motivation
Plant surveys are commonly used to evaluate the performance of mineral processing circuits and to calibrate process simulation models. Historically, survey analysis has been performed using a two-stage workflow.
In the traditional approach, plant survey measurements are first adjusted using data reconciliation techniques so that the flowsheet satisfies mass balance constraints. The reconciliation problem is typically formulated as a weighted least-squares adjustment in which measured values are modified within their estimated uncertainties while enforcing the linear mass balance equations for the flowsheet. The result is a reconciled stream table that exactly satisfies material conservation and is treated as the best estimate of the plant state. Model parameters are then calibrated in a second step so that the process model reproduces the reconciled data.
The mass reconciliation methodology originated at a time when detailed flowsheet simulation models were not widely available. The linear structure of the mass balance equations allowed plant survey data to be analysed using relatively simple optimisation methods, making reconciliation a practical tool for interpreting plant measurements. Process model parameter calibration was later incorporated as a second step in which models were fitted to the reconciled stream table.
Modern process simulators now incorporate detailed models of equipment behaviour and circuit interactions. Grinding and classification circuits, for example, are governed by population balance breakage models, classification partition functions, and recirculating load relationships. These nonlinear process models already enforce strong physical constraints on the feasible flowsheet solution.
SurveyFit takes advantage of these models by estimating parameters directly from the measured survey data. During the optimisation process the flowsheet simulator itself enforces mass conservation and equipment behaviour while parameters are adjusted to minimise the difference between simulated and measured values.
In this formulation the process model effectively replaces the traditional reconciliation step. The optimisation searches for the parameter set that produces a flowsheet solution consistent with both the measurements and the physical constraints of the circuit.
Simultaneous parameter estimation using a strongly constrained flowsheet model therefore provides a modern alternative to the traditional two-step reconciliation workflow for grinding circuit survey analysis. While this approach is not universally superior in all contexts, it is a practical and often simpler approach in situations where the underlying process physics are well represented by the simulation model.
Applications
SurveyFit is intended for calibration of steady-state process flowsheets using plant survey data. Typical mineral processing applications include grinding circuits, flotation circuits, gravity concentration systems, dense medium separation circuits, and other beneficiation flowsheets.
The same approach may also be applied to other types of process flowsheets modelled in SysCAD, including hydrometallurgical circuits, pyrometallurgical processes, chemical refining systems, and other metallurgical or chemical processing operations.
Measurements used during fitting may include plant survey data such as particle size distributions, assay grades, and solids fractions. Instrument measurements such as flow, density, pressure, and equipment power draw may also be incorporated.
In practice, any process variable represented by a SysCAD tag can be included in the regression provided that a corresponding measured value is available. This allows measurements from multiple unit operations across an entire flowsheet to be incorporated into a single calibration problem.
Key Features
SurveyFit provides a set of capabilities specifically designed for calibration of process simulation models using plant survey data. These features support the definition of calibration problems, execution of parameter estimation, and interpretation of the resulting model fit.
Key features include:
- Direct parameter estimation for process simulation models
- Integration with SysCAD steady-state flowsheets
- Calibration of complete flowsheets including recirculating circuits
- Parameter bounds for controlled regression
- Flexible handling of survey measurements and measurement uncertainty
- Graphical diagnostics including parity plots and residual analysis
Installation and Licensing
This section describes the procedures for installing and licensing the SurveyFit software on a client PC.
SurveyFit is distributed as a portable Windows application and does not require a formal installer. Installation therefore consists of extracting the distribution archive to a suitable folder and ensuring that SysCAD is available on the system.
Installation
SurveyFit may be installed by following the procedures below:
- The software is distributed as a ZIP archive containing the application and supporting files.
- Installation consists of extracting the archive to a suitable folder.
- Administrator rights are not normally required to install or use the software.
Some basic knowledge of file manipulation in Microsoft Windows is required to complete the installation.
System requirements
The following hardware and software are required to use SurveyFit:
- Microsoft Windows 10 or later
- A licensed installation of SysCAD with COM automation enabled
The SysCAD flowsheet used during calibration must be able to solve successfully in steady-state operation before using SurveyFit.
Step One: Obtain the SurveyFit software
SurveyFit is distributed as a ZIP archive containing the application and example files.
You can obtain the distribution package through any of the following methods:
- Download the archive from the Met Dynamics customer downloads page.
- Request the latest version from Met Dynamics by contacting support@metdynamics.com.au.
- Obtain the version bundled with certain SysCAD distributions where applicable.
Step Two: Extract the software
Extract the contents of the SurveyFit distribution archive to a folder on the local system.
The recommended installation location is: C:\Met Dynamics\SurveyFit
This allows SurveyFit to be installed alongside other Met Dynamics software tools, although any suitable directory may be used.
The installation folder should contain the SurveyFit executable and supporting files.
Step Three: Verify SysCAD connectivity
SurveyFit communicates with SysCAD using the SysCAD COM Automation interface.
Before using SurveyFit, confirm the following:
- SysCAD is installed and licensed on the machine.
- The flowsheet model to be calibrated solves successfully in steady-state mode.
- SysCAD is registered as a COM application on the system.
Instructions for registering SysCAD for COM automation are available in the SysCAD documentation on COM Automation.
SurveyFit will connect to SysCAD automatically when a calibration solve is executed.
Licensing
SurveyFit licenses are provided on a per-seat basis. Each license is locked to the hardware identifier of an individual client PC.
If no valid license is installed, SurveyFit operates in demonstration mode. Detailed information on the License dialog and Demo Mode behaviour is provided on the Reference page.
Obtaining a license
To obtain a license for SurveyFit:
- Start the SurveyFit application.
- Open the Help → License → Show Details dialog.
- Copy the Site Code displayed in the dialog window.
- Paste the Site Code into an email message.
- Include your contact details and request a license.
- Send the request to accounts@metdynamics.com.au.
After the request has been received, Met Dynamics will generate a license file and return it by email.
Installing the license
Once the license file has been received:
- Save the license file (for example
MetDynamics.lic) to the SurveyFit installation folder. - Restart the SurveyFit application.
- Open the Help → License dialog.
- Confirm that the License status field reports LICENSE_OK.
The software is now fully licensed.
This procedure only needs to be performed once for a given machine.
Trial licenses
Trial licenses may be issued upon request.
Trial licenses:
- operate for a limited evaluation period
- provide full software functionality during the trial period
Please contact support@metdynamics.com.au to request a trial license.
Uninstallation
SurveyFit may be removed from the system by deleting the installation folder containing the application files.
Configuration settings stored for the current user may remain in the Windows registry after removal of the application files.
Further support
If any step of the installation or licensing procedures fails, please contact support@metdynamics.com.au for assistance.
User Interface
SurveyFit uses a tab-based interface organised around the calibration workflow. Each tab corresponds to a stage of the parameter estimation process, from defining adjustable parameters and survey measurements through to executing the regression and analysing the results. Detailed descriptions of fields, table columns, menus, dialogs, and solver options are provided on the Reference page.
Parameters tab
The Parameters tab defines the model parameters that may be adjusted during regression. Each parameter corresponds to a SysCAD tag representing a model variable.
Lower and upper bounds constrain the allowable parameter range.
Measurements tab
The Measurements tab defines the plant measurements used during parameter estimation.
Each measurement corresponds to a SysCAD tag and includes the measured value, an error model, and a standard deviation representing measurement uncertainty.
Particle size distribution measurements may optionally use the Whiten PSD error model.
Solve tab
The Solve tab controls execution of the parameter estimation solve.
During regression SurveyFit maintains a SysCAD simulation session and repeatedly evaluates the flowsheet using updated parameter values. The objective history plot shows the evolution of the regression objective during the solution process.
Results tab
The Results tab provides graphical diagnostics for evaluating the quality of the parameter estimation. These diagnostics include parity plots comparing measured and estimated values together with residual distributions.
Issues panel
The Issues panel is a docked panel that displays configuration problems detected in the current SurveyFit project. These messages highlight issues that may prevent the regression from running correctly, such as missing measurements, invalid parameter bounds, or incomplete configuration.
Log panel
The Log panel is a docked panel that displays execution messages generated during flowsheet evaluation and parameter estimation. The log provides diagnostic information during regression and may assist in troubleshooting solver behaviour or communication with SysCAD.
Project Files
SurveyFit projects are stored using the .sfit file format. A project file contains all information required to define and reproduce a calibration problem. Further details are provided on the Reference page.
A project typically contains:
- parameter definitions
- measurement definitions
- solver configuration
- regression results
Project files allow calibration problems to be saved, reopened, and refined as additional measurements become available or as the flowsheet model evolves.
In practice, a SurveyFit project is created and progressively configured as part of the calibration workflow described below.
Typical Workflow
A typical workflow when using SurveyFit is outlined below. Detailed descriptions of the related tabs, controls, and dialogs are provided on the Reference page.
- 1. Prepare a SysCAD model
- Prepare a SysCAD flowsheet representing the process to be calibrated. The model should be able to solve successfully in steady state before attempting parameter estimation.
- 2. Create or open a SurveyFit project
- Launch SurveyFit and create a new project or open an existing one. The project stores all information required for the calibration problem.
- SurveyFit connects to the selected SysCAD project and maintains the simulation session during the fitting process.
- 3. Import measurements
- Define the measured plant data to be used during regression. Measurements may be imported from external data sources or entered directly within the application.
- 4. Select adjustable parameters
- Choose which model parameters will be adjusted during fitting. These parameters correspond to SysCAD tag values representing model variables.
- 5. Configure regression settings
- Specify parameter bounds, measurement standard deviations, and other configuration options that influence the regression behaviour.
- 6. Run regression
- Execute the regression procedure. SurveyFit repeatedly updates model parameters, solves the SysCAD flowsheet, and evaluates the difference between measured and simulated values.
- 7. Analyse results
- Examine parity plots, residual diagnostics, and fitted parameter values to assess the quality of the regression.
- 8. Modify configuration and re-run regression
- If required, parameters, measurement weights, or model configuration may be adjusted before running the regression again. The SysCAD simulation session remains active during this process.
- 9. Save results back to SysCAD
- Once a satisfactory fit is achieved, the final parameter values can optionally be written back to the SysCAD project. The calibration configuration and results may also be saved to a SurveyFit project file for future use.
Mathematical Formulation
The parameter estimation problem is formulated as a nonlinear regression in which the difference between measured plant data and model predictions is minimised.
- [math]\displaystyle{ S(\theta)=\sum_{i=1}^{N} \left( \frac{y_i^{\mathrm{sim}}(\theta)-y_i^{\mathrm{meas}}} {\sigma_i} \right)^2 }[/math]
where:
- [math]\displaystyle{ \theta }[/math] is the vector of adjustable model parameters
- [math]\displaystyle{ y_i^{\mathrm{sim}}(\theta) }[/math] is the simulated value predicted by the flowsheet model
- [math]\displaystyle{ y_i^{\mathrm{meas}} }[/math] is the measured survey value
- [math]\displaystyle{ \sigma_i }[/math] is the measurement standard deviation
- [math]\displaystyle{ N }[/math] is the number of measurements
Measurements with smaller standard deviations therefore exert greater influence on the regression.
Model Preparation Guidelines
Reliable calibration requires that the SysCAD flowsheet model behaves robustly during repeated evaluation.
Use a simplified flowsheet
- Parameter estimation is generally easier when the model used for fitting is kept as simple as possible. Additional model complexity can be introduced after the primary model parameters have been calibrated.
- Where possible, remove equipment that does not influence the survey measurements or replace complex downstream sections with simplified boundary conditions.
Avoid discontinuities
- Discrete switching logic or other discontinuous model behaviour can make regression unstable. Optimisation algorithms assume that model outputs change smoothly with parameter variation. Sudden discontinuities may therefore prevent the solver from converging.
- Continuous model behaviour is generally preferred during parameter estimation.
Avoid PID controllers
- Closed-loop control elements such as PID controllers should be avoided during parameter estimation. Active controllers continuously adjust manipulated variables while the regression algorithm is simultaneously adjusting model parameters, introducing additional nonlinear feedback into the flowsheet and increasing SysCAD solve time. Because the flowsheet must be solved many times during optimisation, the presence of active control loops can significantly increase the overall time required for parameter estimation.
Ensure stable solver convergence
- SurveyFit repeatedly solves the SysCAD flowsheet during the regression process. Models that do not converge reliably may significantly reduce regression performance.
- If the model is unlikely to converge for a given parameter set, it is generally preferable for the flowsheet to fail quickly rather than consume excessive solver iterations.
Use realistic initial parameter values
- Parameter estimation generally converges faster and more reliably when the starting parameter values are already physically reasonable. Where possible, parameters should be set to typical values based on engineering judgement before running the regression.
- Previous calibrated survey models can provide very good initial values when a similar circuit or ore type has been modelled before. Standalone calibrated unit models may also provide useful starting values, even if they were developed outside the full flowsheet environment, for example using the Met Dynamics Models Excel Add-In or other single-unit calibration tools.
- In practice, the user should aim to begin from a flowsheet state that is already approximately correct. For example, solids flow rates, water balances, circulating loads, and major stream conditions should be broadly consistent with the surveyed plant. Starting from a roughly correct solution can significantly reduce solve time and improve the likelihood of successful convergence.
Watch for recirculating load runaway
- Certain parameter combinations may lead to rapidly increasing recirculating loads in recycle streams. When this occurs the flowsheet may fail to converge or require excessive solver iterations.
- If this behaviour is observed, parameter bounds may need to be tightened or the parameter set adjusted to prevent unstable operating regions.
Fit circuit sections in logical stages
- For complex flowsheets it is often beneficial to calibrate different parts of the circuit separately before performing a full circuit fit.
- For example, a grinding circuit may first be calibrated for the primary SAG mill and pebble crushing section. Once these parameters are stable, the secondary ball mill and cyclone circuit can be fitted. A final full-circuit regression can then be used to refine the overall solution.
- This staged approach can significantly improve convergence and reduce regression time compared with fitting all parameters simultaneously.
Results Interpretation
SurveyFit provides several diagnostic outputs to assist interpretation of the regression results. These diagnostics should be interpreted together when evaluating the quality of a calibration.
These include:
- measured vs estimated parity plots
- distributions of standardised residuals
- summary statistics for the objective function
Detailed descriptions of the corresponding plots, summary fields, and results displays are provided on the Reference page.
Parity plots allow rapid visual assessment of model accuracy. Ideally the estimated values should lie close to the parity line, indicating that the calibrated model reproduces the survey measurements without systematic bias.
Systematic deviations from the parity line may indicate model bias or structural issues in the flowsheet model. For example, points consistently above or below the parity line may indicate biased model predictions, while curvature or separation between groups of measurements may suggest that particular unit operations or measurement types are not being reproduced correctly.
Residual distributions provide additional insight into the statistical consistency of the regression. When measurement uncertainties are specified realistically and the model structure is appropriate, the standardised residuals should be approximately centred around zero and display a roughly symmetric distribution.
As a practical guideline, most residuals would typically be expected to lie within approximately ±2 when expressed in units of the measurement standard deviation. Occasional larger residuals may occur, but systematic clustering away from zero or many values exceeding this range may indicate underestimated measurement uncertainty, inconsistent survey measurements, or deficiencies in the flowsheet model.
Large residuals may therefore highlight measurements that require review or aspects of the model that may require further refinement.
Objective statistics summarise the overall goodness of fit across all measurements included in the regression. These statistics should always be interpreted together with engineering judgement regarding the plausibility of the estimated parameter values.
A satisfactory regression result should demonstrate good agreement on parity plots, residuals centred around zero, and parameter values that remain physically realistic.
Examples
A set of example SurveyFit projects is distributed with the software installation. These examples demonstrate calibration workflows for a variety of process flowsheets and survey data sets.
Examples include typical mineral processing circuits such as grinding circuits, flotation circuits, and other beneficiation flowsheets.
Detailed descriptions of the example projects are provided on the SurveyFit (Examples) page.
Next Steps
Once a satisfactory calibration has been obtained, the fitted model can be used for further analysis of the circuit within the SysCAD simulation environment. The calibrated parameter set represents the best estimate of the circuit behaviour consistent with both the plant survey data and the underlying process model.
The calibrated model may then be used to evaluate circuit performance, investigate operating conditions, or assess potential circuit modifications. Because the parameters have been estimated directly from plant survey measurements, the resulting simulation provides a practical basis for engineering analysis and scenario evaluation.