home Forums # Technical Support Mamdani Output

Viewing 2 posts - 1 through 2 (of 2 total)
  • Author
    Posts
  • #1100
    Unknown
    Member

    Hi,
    I’m having some problem to figure out how the output of a really simple example is computed.

    Below I have attached the output of my model.
    While I’m completely ok with the two “Accumulating terms”:
    – Accumulating term: UP Thresholded 0.600 Minimum term: UP
    – Accumulating term: STY Thresholded 0.133 Minimum term: STY

    I cannot figure out why these terms becames 0.541 (instead of 0.6) and 0.094 (instead of 0.133):

    DeltaC.fuzzy = 0.000/BDW + 0.000/DWN + 0.094/STY + 0.541/UP + 0.000/BUP

    Of course I am missing something in the fuzzy theory, and for this I just need link to know more about these values. Otherwise, it could be a bug but I do not think so !

    The engine is set as:
    – conj: Min
    – disj: Max
    – activation: Min
    – accumulation: Max

    Thanks,
    Massimo
    (below the fuzzylite output)

    Engine type: Mamdani
    /src/Engine.cpp::process[205]:===============
    /src/Engine.cpp::process[206]:CURRENT INPUTS:
    /src/Engine.cpp::process[212]:Cres.input = 0.120
    /src/Engine.cpp::process[213]:Cres.fuzzy = 0.600/LOW + 0.133/FINE + 0.000/HIGH
    /src/Engine.cpp::process[212]:Rgain.input = 0.200
    /src/Engine.cpp::process[213]:Rgain.fuzzy = 0.000/LOW + 1.000/FINE + 0.000/HIGH
    /src/rule/RuleBlock.cpp::activate[47]:===================
    /src/rule/RuleBlock.cpp::activate[48]:ACTIVATING RULEBLOCK
    /src/rule/RuleBlock.cpp::activate[51]:if Cres is LOW and Rgain is LOW  then DeltaC is BUP [activationDegree=0]
    /src/rule/RuleBlock.cpp::activate[51]:if Cres is LOW and Rgain is FINE  then DeltaC is UP [activationDegree=0.6]
    /src/rule/Consequent.cpp::modify[74]:Accumulating term: UP Thresholded 0.600 Minimum term: UP Triangle 0.050 0.125 0.200
    /src/rule/RuleBlock.cpp::activate[51]:if Cres is LOW and Rgain is HIGH  then DeltaC is UP [activationDegree=0]
    /src/rule/RuleBlock.cpp::activate[51]:if Cres is FINE and Rgain is LOW  then DeltaC is UP [activationDegree=0]
    /src/rule/RuleBlock.cpp::activate[51]:if Cres is FINE and Rgain is FINE  then DeltaC is STY [activationDegree=0.133333]
    /src/rule/Consequent.cpp::modify[74]:Accumulating term: STY Thresholded 0.133 Minimum term: STY Triangle -0.100 0.000 0.100
    /src/rule/RuleBlock.cpp::activate[51]:if Cres is FINE and Rgain is HIGH  then DeltaC is DWN [activationDegree=0]
    /src/rule/RuleBlock.cpp::activate[51]:if Cres is HIGH and Rgain is LOW  then DeltaC is STY [activationDegree=0]
    /src/rule/RuleBlock.cpp::activate[51]:if Cres is HIGH and Rgain is FINE  then DeltaC is DWN [activationDegree=0]
    /src/rule/RuleBlock.cpp::activate[51]:if Cres is HIGH and Rgain is HIGH  then DeltaC is BDW [activationDegree=0]
    /src/Engine.cpp::process[229]:===============
    /src/Engine.cpp::process[230]:CURRENT OUTPUTS:
    /src/Engine.cpp::process[235]:DeltaC.default = nan
    /src/Engine.cpp::process[238]:DeltaC.lockRange = 0
    /src/Engine.cpp::process[241]:DeltaC.lockValid = 0
    /src/Engine.cpp::process[246]:DeltaC.output = 0.0905711
    /src/Engine.cpp::process[248]:DeltaC.fuzzy = 0.000/BDW + 0.000/DWN + 0.094/STY + 0.541/UP + 0.000/BUP
    /src/Engine.cpp::process[249]:term: fuzzyOutput Accumulated -1.000 1.000 Maximum term: UP Thresholded 0.600 Minimum term: UP Triangle 0.050 0.125 0.200 term: STY Thresholded 0.133 Minimum term: STY Triangle -0.100 0.000 0.100
    /src/Engine.cpp::process[254]:============== 
    #1101

    Hi Massimo,

    thank you for your post and for noticing that tricky bit.

    The fuzzy output is indeed what you expect, but what you see is not a bug. The “fuzzy output” shown in QtFuzzyLite v4.0 refers to the membership value of the defuzzified value over the terms in the variable, that is, you are seeing the fuzzified output value (\mu(y)). I have addressed this bit in QtFuzzyLite v5.0 by showing both the fuzzy output value that you refer to (\tilde{y}), and the fuzzified output value (\mu(y)). QtFuzzyLite v5.0 will be released early July, 2014, and it will be truly amazing. Stay tuned!

    I have mentioned this bit in the forum of known issues:

    Fuzzy Output Text in qtfuzzylite

    Let me know if that answers your question.

    Cheers,

    Juan

Viewing 2 posts - 1 through 2 (of 2 total)
  • You must be logged in to reply to this topic.