single water tank using Fuzzy PID

[Whc]

i'm doing a prototype project. Do you guys know how to program the fuzzy PID to the water flow to the tank depend to the water level in the tank. Is urgent. Can u guys give me a similar coding. so that i can study. Thx

[Whc]

// Define rule base
struct rule_set {
float lowlimit; // low limit, membership 0.0
float highlimit; // high limit, membership 1.0
float weighting; // weighting factor for inference combination
} ;
typedef struct rule_set rule;

// Membership bounds on input and output levels
#define P_LIMIT 20000.0
#define I_LIMIT 400000.0
#define D_LIMIT 5000.0
#define O_LIMIT 20000.0

// Weighting factors for inference
#define P_RULE_WEIGHT 10.000
#define I_RULE_WEIGHT 0.040
#define D_RULE_WEIGHT 0.120

static rule P_rule = { -P_LIMIT, P_LIMIT, P_RULE_WEIGHT };
static rule I_rule = { -I_LIMIT, I_LIMIT, I_RULE_WEIGHT };
static rule D_rule = { -D_LIMIT, D_LIMIT, D_RULE_WEIGHT };
static rule O_rule = { -O_LIMIT, O_LIMIT, 1.0f };

// Evaluate "is member" and "is not member" of input value, given rule
// to apply. Accumulates results with previous results. Takes advantage
// of special complementation properties of this rule set.
void evaluate_support(
float variable, rule *pRule, float *isMember, float *isNotMember )
{
float in_set;
if ( variable >= pRule->highlimit )
in_set = 1.0;
else if ( variable <= pRule->lowlimit )
in_set = 0.0;
else
in_set = (variable - pRule->lowlimit) /
(pRule->highlimit - pRule->lowlimit);
*isMember += in_set * pRule->weighting ;
*isNotMember += (1.0 - in_set) * pRule->weighting ;
}

// Normalize and evaluate output
float defuzzify( rule *pRule, float isMember, float isNotMember )
{
float output;
output = (isMember * pRule->highlimit) + (isNotMember * pRule->lowlimit);
if (output>pRule->highlimit) output=pRule->highlimit;
if (output>pRule->lowlimit) output=pRule->lowlimit;
return output;
}

float compute_fuzzy_PID ( float Err, float AccumErr, float Speed )
{
float isMember = 0.0;
float isNotMember = 0.0;
evaluate_support( Err, &P_rule, &isMember, &isNotMember );
evaluate_support( AccumErr, &I_rule, &isMember, &isNotMember );
evaluate_support( Speed, &D_rule, &isMember, &isNotMember );
return defuzzify( &O_rule, isMember, isNotMember );
}
i get the coding but dont understand.Can someone explain to me?thx.

[temtronic]

I gave up on 'fuzzy' logic over 20 years ago once I found how to get Matlab's 'Simulink' to run in true realtime. Since then I've never seen a fuzzy PID controller that had better response(speed and/or accuracy) that a properly tuned analog PID controller.The test was to control a 3DOF helicopter via the Net in realtime with 2 other applications open and running.

The 'fuzzy' concept is just that 'fuzzy',nothing more than a HUGE collection of if-then-else statements tested one after another after another ad naueum.... Microchip does have some good(?) examples of it
in the embedded applications handbook(1st or 2nd edition ?).

The classic example is in home heating..again..a series of if..then..else statements.Find one of them..your water tank level is very, very similar,easy enough to modify.

[SherpaDoug]

I thought the fuzzy logic fad passed a few years (many software generations) ago. Is there some outdated spec that reqires you to use fuzzy logic? It is like requiring you to use object oriented programming. That fad is getting a little long in the tooth too.
_________________
The search for better is endless. Instead simply find very good and get the job done.

[Whc]

is it use series if else can get similar result with fuzzy PID coding(posted)?