A trade study or trade-off study is the activity of a multidisciplinary team to identify the most balanced technical solutions among a set of proposed viable solutions (System Engineering Manual, Federal Aviation Administration, 2006).  

A trade study is used to compare with a number of alternative solutions to see whether and how well they satisfy a particular set of criteria. Each solution is characterized by a set of measures of effectiveness (often abbreviated “moe’s”) that corresponds to evaluation criteria and has a calculated value or value distribution. The moe’s for a given solution is then evaluated using an objective function (often called a cost function or utility function), and the results for each alternative are compared to select a preferred solution.

Magic Model Analyst has built-in support for trade study analysis. The TradeStudyExamples sample model is used as a demonstration for trade study analysis through the following steps.

  1. Use TradeStudy «Block» «Analysis» through the following step:
    • From the MD Customization for SysML::analysis patterns::trade study package, drag TradeStudy «Block» «Analysis» into the Block Definition diagram (BDD).

      Note

      The TradeStudy package is available in all SysML projects. If the MD Customization for SysML package is not visible, click in the Containment tree pane and select Show Auxiliary Resources.

    The TradeStudy Block contains the following four value properties which will be inherited by the TradeAnalysis Block.

    • OutOfSpec: Real: counts how much percent of sample value that violates any attached constraints or Requirements.
    • N: Integer: the number of configurations evaluated.
    • score: Real: set as the winner value of the «objectiveFunction».
    • winner: String: used to keep the winner configuration as a string.
  2. Create the TradeAnalysis Block by doing the following:
    1. Inherit the TradeStudy (Block).
      Create a new Block and Generalization from the TradeStudy Block so that the new TradeAnalysis Block is inherited, e.g., an instance of BrakeTradeStudy. Those four inherited value properties will be assigned to the TradeAnalysis Block and updated real-time during the execution.
    2. Add main simulation context as Part.
      Add the main Block of simulation context as a Part property of the TradeAnalysis Block, e.g., Main : VehicleAnalysis.
    3. Create «objectiveFunction».
      An objective function is a special type of constraint Block used to determine the overall value of any weighted alternatives in terms of weighted criteria. You must create a constraint Block containing related constraint parameters and a constant expression for determining the best weighted value, e.g., Criteria.

      Note

      • The specification of the constraint must be an equation with LHS = RHS, where LHS contains only one parameter to bind with TradeAnalysis::^score, and RHS can contain multiple parameters (bound with corresponding value properties) to evaluate as winner criteria.
      • The output of «objectiveFunction» will be compared with previous weighted results. If an alternative is greater, it will be set as the winner (maximum value by default). However, if you want the minimum value, use a negative value, e.g., value = -distance.
    4. Add «objectiveFunction».
      Create the constraint property in the TradeAnalysis Block typed by the created constraint Block, and apply «objectiveFunction», e.g., OF : Criteria.

    5. Create «alternatives».
      Create reference properties typed by a Block of alternatives and apply «alternatives» (Qualified Name: MD Customization for SysML::additional_stereotypes::alternatives) to those properties, e.g., C : Caliper, R : Rotor, and P : Pad, as shown in the figure below.


      Structure of the TradeAnalysis Block.
  3. Create binding of the TradeAnalysis Block through the following steps:
    1. Create an Internal Block diagram (IBD).
      Create an IBD of the TradeAnalysis Block. You must select Parts which have been set as «objectiveFunction» and «alternatives» to display. You must also display ^score (inherited property) in the diagram.
    2. Bind «objectiveFunction».
      You must bind ^score with a Binding Connector to the LHS parameter of «objectiveFunction», e.g., ^score – value. You must also bind a value property of the main simulation context with a Binding Connector to the RHS parameter of «objectiveFunction», e.g., stoppingDistance – distance.
    3. Bind «alternatives».
      You must bind each «alternatives» with a Binding Connector to each Part property.
    4. Set kind/source tags.
      For each «alternatives», source depends on the following kinds through the Tags settings.
      • kind = TABLE: source must be an instance table which must be the same Classifier as the alternative property, as shown in the two figures below. However, sorting of rows in the table is not necessary.

        TABLE, kind of «alternatives», must be the same Classifier as the alternative property.

        Binding of the TradeAnalysis Block in the Internal Block diagram (kind = TABLE).
      • kind = SUBTYPES: source must be a parent Block of subtypes which must be the same type as the alternative property, as shown in the two figures below. Also, the parent Block will not be evaluated as well as any Blocks which have Is Abstract = true.

        SUBTYPES, kind of «alternatives», must be the same type as the alternative property.

        Binding of the TradeAnalysis Block in the Internal Block diagram (kind = SUBTYPES).
      • kind = EXCEL: source must be an instance table linked to an Excel file, and table columns must be mapped to Excel/CSV columns as shown in the two figures below (see also Sync with Excel or CSV files). However, you do not need to use the Read from File command from the Publish Excel toolbar to load data into the table.

        EXCEL, kind of «alternatives», must be linked to an Excel file, and table columns must be mapped to Excel/CSV columns.

        Binding of the TradeAnalysis Block in the Internal Block diagram (kind = EXCEL).
  4. Create SimulationConfig and other settings.
    You can create a SimulationConfig, so that it can specify the resultLocation of this TradeStudy to a package, e.g., instances of the winning configuration will be saved. Also, the following tags can be set, as shown in the figure below.
    1. executionTarget
      executionTarget
      must be set to the TradeAnalysis Block, e.g., BrakeTradeStudy in the sample. executionTarget can be an instance of the TradeAnalysis Block so that the instance can be reconfigured.
    2. resultLocation
      A package/instance table must be specified so an instance with related information will be saved after running the simulation. The information includes N, OutOfSpec, score, winner, and other elements.

      Note

      • If you do not create a SimulationConfig and run the TradeAnalysis Block directly, TradeStudy will not be triggered, but the TradeAnalysis Block will be run as normal .
      • If the executionTarget of a SimulationConfig is the TradeAnalysis Block, TradeStudy execution will override other executions, so Monte Carlo simulation will not be triggered even if numberOfRuns is more than 1.
    3. silent
      You can set the silent tag accordingly to see the animation.
    4. rememberFailureStatus
      You can use rememberFailureStatus when evaluating those weighted alternatives. Any alternatives violating any of the attached constraints/Requirements will not be considered as the winner, but they will be used in the calculation of OutOfSpec.
    5. Tags neglected

      You can neglect animationSpeed, constraintFailureAsBreakpoint, UI, and autoStart.

      SimulationConfig created for other settings, e.g., executionTarget and resultLocation.
  5. Run the SimulationConfig.
    When running the SimulationConfig, the information of TradeStudy will be printed in the Console pane. The Simulation pane will be disabled, but all warning/errors will be printed. There is a progress bar shown with the description of Executing alternative [n] of [total iterations] (Time elapsed: [time]), and the Cancel button that allows canceling Trade Study as shown in the figure below. Simulation Toolkit automatically iterates all variants and instantiates all possible configurations in memory. For example, Brake which contains the combinations of Pad (26 instances), Caliper (20 instances), and Rotor (4 instances) will have 26 x 20 x 4 = 2,080 configurations. The winner value on each iteration will be compared directly with the score value property.

    Note

    • The Starting Math Engine progress bar will be shown in this sample because MATLAB is used as the external evaluator. See also Integration with MATLAB.
    • Any alternatives violating any of the attached constraints/Requirements will not be considered as the winner. They will be used in the calculation of OutOfSpec.
    • rememberFailureStatus of a SimulationConfig will be used when evaluating those alternatives.
    • In some cases, you can click Unlock in the Simulation pane to see execution details during the execution.

    A progress bar shown with description and the Cancel button during the simulation.
  6. Get the TradeAnalysis result.

    When the simulation is either completed or canceled, winning information will be printed on the Console pane in the following three lines as shown in the figure below.

    • The first line shows the number of iterations (completed/canceled) of all alternatives for executionTarget with elapsed time.

    • The second line displays the winning configuration from the winner string.

    • The third line is the winning score from the ^score.

      Note

      The winner string is printed with the formats as follows:
      AlternativeProperty.Name1 : StringKind [, AlternativeProperty.Name2 : StringKind, AlternativeProperty.Name3 : StringKind, …]

      where StringKind will apply the following rules, depending on kind of alternative:

      • kind=TABLE: then StringKind=InstanceName, e.g., P : Saphire 66, C : Alphine K7, R : Rotus 30.

      • kind=SUBTYPES: then StringKind=subtypesName, e.g., power : Diesel, support : Wheels, stopping : Brakes.

      • kind=EXCEL: then StringKind=#Row, e.g., R : #5, P : #21, C : #21, where Row is the number of Excel rows.

    The result instance will be saved at the location as specified in resultLocation of SimulationConfig. You can create an instance table, set a Classifier to the TradeAnalysis Block, and set Scope to the package of the results.

    The TradeAnalysis result (in the last 3 lines) is printed on the Console pane, and the result instance is saved into a package and presented in the instance table.