Simplify: gr.Interface for reliability

app.py

@@ -1,323 +1,125 @@
 """
 Predictive Agent - LSTM-Based RUL Prediction
-HuggingFace Space: davidfertube/predictive-agent
-
-Predicts Remaining Useful Life for CCGT equipment using neural network
-patterns trained on operational health data.
-
+Predicts Remaining Useful Life for CCGT equipment.
 Author: David Fernandez - Industrial AI Engineer
 """
 
 import gradio as gr
-import numpy as np
-
-# ============================================================
-# EQUIPMENT HEALTH PARAMETERS
-# ============================================================
-
-FEATURE_COLUMNS = ['health_index', 'vibration', 'heat_rate_delta',
-                   'operating_hours', 'start_count']
 
+# Thresholds
 THRESHOLDS = {
-    'health_index': {'good': 70, 'warning': 40, 'unit': '%'},
-    'vibration': {'good': 0.3, 'warning': 0.5, 'unit': 'in/s'},
-    'heat_rate_delta': {'good': 4, 'warning': 8, 'unit': '%'},
-    'operating_hours': {'good': 50000, 'warning': 65000, 'unit': 'hrs'},
-    'start_count': {'good': 1000, 'warning': 1200, 'unit': 'starts'}
+    'health_index': (70, 40),      # good, warning (higher is better)
+    'vibration': (0.3, 0.5),       # good, warning (lower is better)
+    'heat_rate_delta': (4, 8),     # good, warning (lower is better)
 }
 
-# ============================================================
-# DEMO DATA - Pre-loaded from ccgt-health-history dataset
-# ============================================================
-
-# Demo 1: Healthy Equipment - Recently overhauled
-DEMO_1_HEALTHY = {
-    'health_index': 96.5,
-    'vibration': 0.14,
-    'heat_rate_delta': 1.2,
-    'operating_hours': 48500,
-    'start_count': 920
-}
-
-# Demo 2: Degraded Equipment - Approaching maintenance
-DEMO_2_DEGRADED = {
-    'health_index': 42.3,
-    'vibration': 0.48,
-    'heat_rate_delta': 8.5,
-    'operating_hours': 68000,
-    'start_count': 1180
-}
+def predict(health_index, vibration, heat_rate_delta, operating_hours, start_count):
+    """Predict Remaining Useful Life."""
 
-# ============================================================
-# RUL PREDICTION MODEL (LSTM-based logic)
-# ============================================================
+    # Normalize factors (0-1 scale)
+    hi_factor = health_index / 100
+    vib_factor = 1 - min(vibration / 1.0, 1)
+    hr_factor = 1 - min(heat_rate_delta / 15, 1)
+    hours_factor = 1 - min(operating_hours / 80000, 1)
+    starts_factor = 1 - min(start_count / 1500, 1)
 
-def predict_rul(health_index, vibration, heat_rate_delta, operating_hours, start_count):
-    """
-    Predict Remaining Useful Life based on current health parameters.
-    Uses weighted factors derived from LSTM model training.
-    """
-    # Normalize inputs
-    hi_factor = health_index / 100  # 0-1 scale
-    vib_factor = 1 - min(vibration / 1.0, 1)  # Lower is better
-    hr_factor = 1 - min(heat_rate_delta / 15, 1)  # Lower is better
-    hours_factor = 1 - min(operating_hours / 80000, 1)  # Lower is better
-    starts_factor = 1 - min(start_count / 1500, 1)  # Lower is better
-
-    # Weighted RUL calculation (based on LSTM feature importance)
-    weights = {
-        'health_index': 0.35,
-        'vibration': 0.25,
-        'heat_rate_delta': 0.20,
-        'operating_hours': 0.12,
-        'start_count': 0.08
-    }
-
-    composite_score = (
-        hi_factor * weights['health_index'] +
-        vib_factor * weights['vibration'] +
-        hr_factor * weights['heat_rate_delta'] +
-        hours_factor * weights['operating_hours'] +
-        starts_factor * weights['start_count']
-    )
-
-    # Convert to RUL cycles (max 200)
-    rul_cycles = int(composite_score * 200)
-
-    # Add some variance for realism
-    rul_cycles = max(0, rul_cycles + np.random.randint(-5, 5))
-
-    return rul_cycles, composite_score
-
-def get_status(value, param):
-    """Get status based on thresholds."""
-    t = THRESHOLDS[param]
-    if param == 'health_index':
-        if value >= t['good']:
-            return 'OK'
-        elif value >= t['warning']:
-            return 'WARNING'
-        return 'CRITICAL'
-    else:
-        if value <= t['good']:
-            return 'OK'
-        elif value <= t['warning']:
-            return 'WARNING'
-        return 'CRITICAL'
+    # Weighted RUL calculation
+    composite = (hi_factor * 0.35 + vib_factor * 0.25 + hr_factor * 0.20 +
+                 hours_factor * 0.12 + starts_factor * 0.08)
 
-def analyze_equipment(health_index, vibration, heat_rate_delta, operating_hours, start_count):
-    """
-    Analyze equipment health and predict RUL.
-    Returns formatted report with recommendations.
-    """
-    rul_cycles, composite_score = predict_rul(
-        health_index, vibration, heat_rate_delta, operating_hours, start_count
-    )
+    rul = int(composite * 200)
 
-    # Determine urgency
-    if rul_cycles < 30:
-        urgency = "IMMEDIATE"
-        urgency_desc = "Schedule emergency maintenance within 48 hours"
-        urgency_color = "red"
-    elif rul_cycles < 100:
+    # Urgency
+    if rul < 30:
+        urgency = "CRITICAL"
+        action = "Schedule emergency maintenance within 48 hours"
+    elif rul < 100:
         urgency = "SCHEDULED"
-        urgency_desc = "Plan maintenance in next available outage window"
-        urgency_color = "orange"
+        action = "Plan maintenance in next available window"
     else:
         urgency = "ROUTINE"
-        urgency_desc = "Continue normal monitoring schedule"
-        urgency_color = "green"
+        action = "Continue normal monitoring"
 
-    # Build status table
-    params = {
-        'health_index': health_index,
-        'vibration': vibration,
-        'heat_rate_delta': heat_rate_delta,
-        'operating_hours': operating_hours,
-        'start_count': start_count
-    }
+    # Status checks
+    def status(val, good, warn, lower_is_better=True):
+        if lower_is_better:
+            return "OK" if val <= good else "WARNING" if val <= warn else "CRITICAL"
+        return "OK" if val >= good else "WARNING" if val >= warn else "CRITICAL"
 
-    status_rows = []
-    for param, value in params.items():
-        status = get_status(value, param)
-        unit = THRESHOLDS[param]['unit']
-        display_name = param.replace('_', ' ').title()
-        status_rows.append(f"| {display_name} | {value:,.1f} {unit} | {status} |")
+    hi_status = status(health_index, 70, 40, lower_is_better=False)
+    vib_status = status(vibration, 0.3, 0.5)
+    hr_status = status(heat_rate_delta, 4, 8)
 
-    status_table = "\n".join(status_rows)
-
-    # Generate recommendations based on degradation factors
-    recommendations = []
+    # Recommendations
+    recs = []
     if health_index < 60:
-        recommendations.append("1. **Hot Gas Path Inspection** - Health index indicates significant degradation")
+        recs.append("1. **Hot Gas Path Inspection** - Health index degraded")
     if vibration > 0.4:
-        recommendations.append("2. **Bearing Analysis** - Elevated vibration requires investigation")
+        recs.append("2. **Bearing Analysis** - Elevated vibration")
     if heat_rate_delta > 6:
-        recommendations.append("3. **Compressor Wash** - Heat rate deviation suggests fouling")
+        recs.append("3. **Compressor Wash** - Heat rate deviation")
     if operating_hours > 60000:
-        recommendations.append("4. **Major Overhaul Planning** - Approaching inspection interval")
-    if start_count > 1100:
-        recommendations.append("5. **Start-based Maintenance** - Review start-related wear items")
-
-    if not recommendations:
-        recommendations.append("Continue normal condition monitoring")
-
-    recs_md = "\n".join(recommendations)
-
-    # Build confidence metrics
-    confidence = int(composite_score * 100)
+        recs.append("4. **Major Overhaul Planning** - High operating hours")
+    if not recs:
+        recs.append("Continue normal condition monitoring")
 
-    return f"""
-# RUL Prediction Report
+    return f"""# RUL Prediction
 
-## Predicted Remaining Useful Life
-# {rul_cycles} cycles
+## Remaining Useful Life: **{rul} cycles**
 
 ## Urgency: {urgency}
-{urgency_desc}
-
----
-
-## Equipment Health Status
-
-| Parameter | Current Value | Status |
-|-----------|---------------|--------|
-{status_table}
-
----
-
-## Model Confidence
-
-**Composite Health Score**: {confidence}%
-
-Based on LSTM pattern analysis trained on 35+ equipment health histories.
-
----
-
-## Maintenance Recommendations
-
-{recs_md}
+{action}
 
 ---
 
-## Next Steps
+## Equipment Status
+| Parameter | Value | Status |
+|-----------|-------|--------|
+| Health Index | {health_index}% | {hi_status} |
+| Vibration | {vibration} in/s | {vib_status} |
+| Heat Rate Delta | {heat_rate_delta}% | {hr_status} |
+| Operating Hours | {int(operating_hours):,} | - |
+| Start Count | {int(start_count):,} | - |
 
-{"**ACTION REQUIRED**: Contact maintenance planner to schedule emergency outage." if urgency == "IMMEDIATE" else "**PLAN AHEAD**: Coordinate with operations for next maintenance window." if urgency == "SCHEDULED" else "**MONITOR**: Review trends at next weekly reliability meeting."}
+## Recommendations
+{chr(10).join(recs)}
 
 ---
-**Model**: [rul-predictor-ccgt](https://huggingface.co/davidfertube/rul-predictor-ccgt) (LSTM Regression)
-**Dataset**: [ccgt-health-history](https://huggingface.co/datasets/davidfertube/ccgt-health-history)
+*Model: [rul-predictor-ccgt](https://huggingface.co/davidfertube/rul-predictor-ccgt)*
 """
 
-def analyze_csv(file):
-    """Analyze uploaded CSV file with health data."""
-    if file is None:
-        return "Please upload a CSV file with equipment health data."
-
-    try:
-        import pandas as pd
-        df = pd.read_csv(file.name)
-
-        # Check for required columns
-        missing = set(FEATURE_COLUMNS) - set(df.columns)
-        if missing:
-            return f"Missing columns: {', '.join(missing)}\n\nExpected: {', '.join(FEATURE_COLUMNS)}"
-
-        # Analyze last row (most recent reading)
-        latest = df[FEATURE_COLUMNS].iloc[-1]
-        return analyze_equipment(*[latest[col] for col in FEATURE_COLUMNS])
-    except ImportError:
-        return "CSV analysis requires pandas. Please enter values manually."
-    except Exception as e:
-        return f"Error processing file: {str(e)}"
-
-def load_demo_1():
-    return [DEMO_1_HEALTHY[col] for col in FEATURE_COLUMNS]
-
-def load_demo_2():
-    return [DEMO_2_DEGRADED[col] for col in FEATURE_COLUMNS]
-
-# ============================================================
-# GRADIO INTERFACE
-# ============================================================
-
-with gr.Blocks(theme=gr.themes.Soft(), title="Predictive Agent") as demo:
-
-    gr.Markdown("""
-    # Predictive Agent
-    **LSTM-Based RUL Prediction for CCGT Equipment**
-
-    Predict Remaining Useful Life for Combined Cycle Gas Turbine equipment.
-    Click a demo below to see immediate results, or enter your own equipment data.
-
-    ---
-    """)
-
-    with gr.Row():
-        demo1_btn = gr.Button("Demo: Healthy Equipment", variant="secondary", size="lg")
-        demo2_btn = gr.Button("Demo: Degraded Equipment", variant="primary", size="lg")
-
-    gr.Markdown("---")
-
-    with gr.Row():
-        with gr.Column(scale=1):
-            gr.Markdown("### Equipment Health Inputs")
-
-            health_index = gr.Number(label="Health Index (%)", value=85.0, minimum=0, maximum=100)
-            vibration = gr.Number(label="Vibration (in/s)", value=0.18, minimum=0, maximum=1)
-            heat_rate_delta = gr.Number(label="Heat Rate Delta (%)", value=2.5, minimum=0, maximum=15)
-            operating_hours = gr.Number(label="Operating Hours", value=52000)
-            start_count = gr.Number(label="Start Count", value=950)
-
-            predict_btn = gr.Button("Predict RUL", variant="primary", size="lg")
-
-        with gr.Column(scale=1):
-            gr.Markdown("### Prediction Results")
-            output = gr.Markdown()
-
-    with gr.Accordion("Upload Health History CSV", open=False):
-        file_upload = gr.File(label="Upload Equipment Health CSV", file_types=[".csv"])
-        upload_btn = gr.Button("Analyze Upload")
-        gr.Markdown("""
-        **Expected columns**: health_index, vibration, heat_rate_delta, operating_hours, start_count
-
-        For best results, upload historical data. Latest row will be analyzed.
-        """)
-
-    gr.Markdown("""
-    ---
-    ### How It Works
-    ```
-    Health Metrics -> Sequence Generation -> LSTM Inference -> RUL Estimate -> Maintenance Plan
-    ```
-
-    **Key Metrics**: Health Index (overall condition), Vibration (mechanical state),
-    Heat Rate Delta (thermal efficiency), Operating Hours, Start Cycles
-
-    **Resources**: [Model](https://huggingface.co/davidfertube/rul-predictor-ccgt) |
-    [Dataset](https://huggingface.co/datasets/davidfertube/ccgt-health-history) |
-    [GitHub](https://github.com/davidfertube/predictive-agent) |
-    [Portfolio](https://davidfernandez.dev)
-
-    *Built by David Fernandez - Industrial AI Engineer*
-    """)
-
-    # Event handlers
-    predict_btn.click(
-        fn=analyze_equipment,
-        inputs=[health_index, vibration, heat_rate_delta, operating_hours, start_count],
-        outputs=output,
-        api_name="predict"
-    )
-
-    demo1_btn.click(fn=load_demo_1, outputs=[health_index, vibration, heat_rate_delta,
-                                              operating_hours, start_count],
-                    api_name="demo_healthy")
-    demo2_btn.click(fn=load_demo_2, outputs=[health_index, vibration, heat_rate_delta,
-                                              operating_hours, start_count],
-                    api_name="demo_degraded")
-
-    upload_btn.click(fn=analyze_csv, inputs=file_upload, outputs=output, api_name="analyze_csv")
+demo = gr.Interface(
+    fn=predict,
+    inputs=[
+        gr.Number(label="Health Index (%)", value=85, minimum=0, maximum=100),
+        gr.Number(label="Vibration (in/s)", value=0.18, minimum=0, maximum=1),
+        gr.Number(label="Heat Rate Delta (%)", value=2.5, minimum=0, maximum=15),
+        gr.Number(label="Operating Hours", value=52000),
+        gr.Number(label="Start Count", value=950),
+    ],
+    outputs=gr.Markdown(),
+    title="Predictive Agent",
+    description="""**LSTM-Based RUL Prediction for CCGT Equipment**
+
+Predict Remaining Useful Life to optimize maintenance scheduling.
+
+Click an example below to test.""",
+    examples=[
+        [96.5, 0.14, 1.2, 48500, 920],   # Healthy
+        [42.3, 0.48, 8.5, 68000, 1180],  # Degraded
+    ],
+    cache_examples=False,
+    article="""
+## How It Works
+```
+Health Metrics → LSTM Inference → RUL Estimate → Maintenance Plan
+```
+
+**Resources**: [Model](https://huggingface.co/davidfertube/rul-predictor-ccgt) | [Dataset](https://huggingface.co/datasets/davidfertube/ccgt-health-history) | [Portfolio](https://davidfernandez.dev)
+
+*Built by David Fernandez - Industrial AI Engineer*
+"""
+)
 
-demo.queue()
-demo.launch()
+if __name__ == "__main__":
+    demo.launch()