Relationship Analysis between Industrial Control Systems and Data Insight Platform
This document systematically explains the positioning and differences of industrial control systems such as PLC, DCS, and APC, as well as the role of the StarWay Data Insight (StarWayDI) in the Offline Optimization (OO) process. It also looks forward to the key role of the upcoming StarWay Data Online Insight Platform (StarWayODI, essentially an RTO real-time optimization system) in future control loops.
I. Overview of Industrial Control Systems
1.1 Industrial Control Pyramid (ISA-95 Standard Reference)
To understand industrial control, we first look at the "pyramid". Data and control commands flow up and down this pyramid from the underlying hardware devices to the top-level enterprise management:
1.2 Core Functions of Each Layer
| Layer | System Type | Time Scale | Core Tasks |
|---|---|---|---|
| L4 Enterprise Layer | ERP/MES | Day/Week/Month | Production planning, material management, quality traceability |
| L3 Optimization Layer | APC/RTO | Minute/Hour | Multivariable coordination, economic optimization, constraint pushing |
| L2 Control Layer | DCS/PLC | Millisecond/Second | Single-loop control, logical interlock, safety protection |
| L1 Field Layer | Sensors/Actuators | Millisecond | Signal acquisition, action execution |
1.3 Positioning of Data Insight Platform
Analysis:
- StarWayDI (Offline Platform): It is your "offline laboratory". Here you import historical data, use PCA to find anomalies, use PLS to find patterns, and finally get a model (such as knowing that temperature must be between 80-85℃ and pressure between 1.2-1.5 MPa to achieve optimal yield).
- StarWayODI (Under Construction, essentially APC/RTO): It is the "online brain" in the future product line. It receives real-time data from the underlying DCS, uses models trained by StarWayDI for online inference calculations, obtains optimal parameters under current working conditions, and sends these optimization commands (Setpoints) directly to the OPC Server, DCS, or PLC. Note: The dashed frame in the figure indicates that this system is currently under construction planning.
- OPC Server / DCS / PLC (Basic Control & Communication Layer): Receives the setpoints from StarWayODI and executes physical actions such as valve openings and pump speeds.
- DCS / PLC (Basic Control): Execute specific operations such as valve opening and pump speed.
II. Detailed Explanation of PLC, DCS, and APC
2.1 PLC (Programmable Logic Controller)
PLC (Programmable Logic Controller) is a digital computing electronic system specially designed for industrial environments.
Core Features:
- High Reliability: Strong anti-interference ability, suitable for harsh industrial environments
- Flexible Programming: Intuitive programming methods such as ladder diagrams and function block diagrams
- Fast Response: Scan cycle can reach millisecond level
- Low Cost: Suitable for small and medium-sized control tasks
Typical Application Scenarios:
| Scenario | Control Task | Common Brands |
|---|---|---|
| Packaging Machinery | Sequence control, position control | Siemens, Mitsubishi, Omron |
| Production Lines | Start-stop control, speed coordination | Rockwell, Schneider |
| Water Treatment | Pump-valve interlock, liquid level control | Siemens, ABB |
| Building Automation | Air conditioning control, lighting control | Honeywell, Johnson Controls |
PLC Limitations:
- ✅ Good at: Single-loop control, sequential logic control, high-speed response, safety interlock
- ❌ Not good at: Multivariable coordination control, complex optimization calculations, large-scale data processing
2.2 DCS (Distributed Control System)
DCS (Distributed Control System) is a computer control system with decentralized control and centralized management.
Core Features:
- Decentralized Control: Control functions distributed to multiple field control stations
- Centralized Management: Operators monitor the entire plant in the central control room
- Redundant Design: Double backup of key components, high reliability
- Open Architecture: Supports multiple communication protocols and third-party equipment
Typical Application Scenarios:
| Industry | Application Characteristics | Mainstream DCS Brands |
|---|---|---|
| Petrochemical | High temperature and pressure, continuous production | Honeywell, Emerson, Yokogawa |
| Power | Large unit coordinated control | Hollysys, Guodian Zhishen |
| Pharmaceutical | Batch control, GMP compliance | Siemens, Rockwell |
| Metallurgy | Large-scale, high reliability | ABB, Siemens |
DCS vs PLC Comparison:
| Comparison Item | DCS | PLC |
|---|---|---|
| System Architecture | Distributed, networked | Centralized or distributed |
| Control Scale | Large-scale (thousands of points) | Small and medium-scale (hundreds of points) |
| Control Type | Mainly continuous process control | Mainly discrete logic control |
| Engineering Cost | High (suitable for large projects) | Low (suitable for small and medium-sized) |
2.3 APC (Advanced Process Control)
APC (Advanced Process Control) is a class of advanced control technologies beyond traditional PID control, with the core being multivariable model predictive control (MPC).
Core Features:
- Multivariable Coordination: Simultaneously handle multiple interrelated control loops
- Model Prediction: Predict future behavior based on dynamic models
- Constraint Handling: Automatically handle operational constraints
- Maintain Steady State: Run steadily under the premise of satisfying constraints
APC Core Algorithm: MPC (Model Predictive Control)
- Predictive Model: Establish input-output dynamic model based on historical data
- Rolling Optimization: Solve open-loop optimization problem in each control cycle
- Feedback Correction: Use actual measured values to correct model predictions
- Implementation of First Step: Only implement the first control action of the optimization results
APC Application Effects:
| Application Scenario | Typical Benefits | Implementation Cycle |
|---|---|---|
| Distillation Column Control | Product purity increased by 5-10%, energy consumption reduced by 3-5% | 2-3 months |
| Reactor Control | Conversion rate increased by 2-5%, selectivity improved | 3-4 months |
| Furnace Control | Thermal efficiency increased by 2-4%, emissions reduced | 2-3 months |
2.4 Summary of the Three Relationships
III. Offline Optimization (OO) Process
3.1 What is Offline Optimization?
Offline Optimization (OO) refers to the process of analyzing and modeling using historical data to develop optimization strategies without affecting production operations.
Core Features:
- Non-real-time: Based on historical batch data, does not directly control the field
- Safe: Does not affect current production, allows repeated testing
- In-depth: Can perform complex statistical analysis and model verification
- Preparation: Establishes basic models for online optimization (RTO)
3.2 Complete Offline Optimization Process
- Data Preparation (2-4 weeks): Collect historical data, data quality inspection
- Exploratory Analysis (1-2 weeks): PCA analysis, identify abnormal batches
- Modeling Analysis (2-3 weeks): PLS modeling, model verification
- VIP Analysis (1-2 weeks): Identify key variables, develop optimization recommendations
- Verification and Implementation (4-8 weeks): Experimental verification, develop SOP, training implementation
3.3 Model Acceptance Criteria
| Indicator | Minimum Requirement | Good | Excellent |
|---|---|---|---|
| R²Y | > 0.6 | > 0.8 | > 0.9 |
| Q²Y | > 0.5 | > 0.7 | > 0.85 |
| R²Y - Q²Y | < 0.3 | < 0.2 | < 0.1 |
IV. Positioning and Value of StarWayDI
4.1 What is StarWayDI?
StarWay Data Insight (StarWayDI = StarWay Data Insight) is a software tool specially designed for industrial offline data analysis, with core capabilities built on the PCA/PLS model family.
Core Positioning:
- Not a control system, but an analysis tool
- Does not directly control field equipment
- Does not replace DCS/PLC/APC
- Provides data support for optimization decisions
Core Value: -挖掘优化机会 from massive historical data
- Establish quantitative relationship between "process parameters → product quality"
- Identify key control points, guide APC implementation
- Prepare basic models for online optimization (RTO)
4.2 Relationship between StarWayDI and APC
Not competition, but complementarity!
| Dimension | StarWayDI | APC |
|---|---|---|
| Positioning | Offline analysis tool | Online control system |
| Data | Historical data | Real-time data |
| Model | Static model (PLS) | Dynamic model (MPC) |
| Output | Analysis reports, optimization recommendations | Real-time control commands |
| Execution | Manual execution | Automatic execution |
Collaboration Process:
- StarWayDI analyzes historical data, identifies key variables
- StarWayDI establishes static PLS model, evaluates feasibility
- Transfer model/variable information to APC implementation team
- APC engineers establish dynamic MPC model, connect DCS control
- After APC commissioning, StarWayDI regularly analyzes operation data, optimizes models
V. From Offline to Online: StarWayODI Outlook
5.1 Why Online Optimization is Needed?
Limitations of Offline Optimization:
- Sudden changes in raw material batches → offline model failure → product quality fluctuations
- Sudden changes in environmental temperature → offline strategy not applicable → manual adjustment required
- Equipment state drift → offline parameters outdated → optimization effect decreased
Value of Online Optimization:
- Real-time perception of working condition changes
- Automatic adjustment of control strategies
- Continuously maintain optimal operating points
- Reduce manual intervention
5.2 Positioning of StarWayODI
StarWay Data Online Insight Platform (StarWayODI = StarWay Online Data Insight) is the online version of StarWayDI, achieving the leap from offline analysis to online optimization.
Evolution Relationship:
| Feature | StarWayDI | StarWayODI |
|---|---|---|
| Data | Historical data | Real-time data access |
| Model | Static model | Dynamic model update |
| Execution | Manual execution of optimization recommendations | Automatic issuance of optimization commands |
| Operation | Regular analysis | Continuous online operation |
5.3 RTO (Real-Time Optimization) Concept
RTO (Real-Time Optimization) is the core of online optimization.
Hierarchical Relationship between RTO and APC:
| Layer | System | Operation Cycle | Core Tasks |
|---|---|---|---|
| RTO | Real-time optimization layer | 15-60 minutes | Economic benefit optimization, setpoint optimization |
| APC | Advanced control layer | 1-5 minutes | Track RTO setpoints, multivariable coordination |
| DCS | Basic control layer | Millisecond-second level | PID control, execute control commands |
Key Understanding:
- RTO decides "what to do" (setpoint optimization)
- APC decides "how to do it" (dynamic tracking of setpoints)
- DCS decides "to do" (execute control)
5.4 StarWayDI → StarWayODI Evolution Path
VI. System Integration Architecture
6.1 Complete System Architecture Diagram
To achieve a complete closed loop from bottom-layer equipment to top-level optimization, modern industrial internet architecture introduces a Data Platform & Communication Layer (including IoT gateways and Time Series Databases). As the RTO/APC control brain, StarWayODI directly relies on this layer for high-speed data read/write and command issuance.
6.2 Data and Command Flow
VII. Implementation Path Recommendations
7.1 Phased Implementation Roadmap
7.2 Key Success Factors
| Phase | Key Success Factors | Risk Points |
|---|---|---|
| Offline Setup | Data quality, engineer training | Data missing, resistance from field staff |
| In-depth Application | Business value verification, management support | Analysis results fail to guide actual production |
| Online Control Upgrade | Model accuracy, OPC communication stability | Dynamic model mismatch, underlying PID loop anomalies |
| Closed-loop & Self-learning | System adaptive capability, network security | System failure under extreme conditions, data security risks |
7.3 Return on Investment Analysis
| Investment Phase | Main Investment Content | Expected Core Benefits | Payback Period |
|---|---|---|---|
| StarWayDI (Offline Phase) | Software license, offline data implementation services | Enhance process cognition, discover optimization potential, 2-5% process optimization | 6-12 months |
| StarWayODI (Online Phase) | Software license, IoT/TSDB hardware & interface development, control engineering services | Achieve closed-loop automatic control, 3-8% energy consumption reduction, yield improvement | 12-18 months |
| Plant-wide Rollout | Multi-unit expansion licenses, continuous maintenance | Comprehensive benefit improvement, moving towards dark factories / unmanned operation | 18-24 months |
Appendix: Terminology Quick Reference
| Term | English Full Name | Chinese Explanation |
|---|---|---|
| PLC | Programmable Logic Controller | 可编程逻辑控制器 |
| DCS | Distributed Control System | 分布式控制系统 |
| APC | Advanced Process Control | 先进过程控制 |
| MPC | Model Predictive Control | 模型预测控制 |
| RTO | Real-Time Optimization | 实时优化 |
| OO | Offline Optimization | 离线优化 |
| MES | Manufacturing Execution System | 制造执行系统 |
| ERP | Enterprise Resource Planning | 企业资源计划 |
| OPC | OLE for Process Control | 过程控制OLE |
| SCADA | Supervisory Control And Data Acquisition | 数据采集与监视控制 |
This document is the technical white paper of the StarWay Data Insight, helping users understand the industrial control system ecosystem and the positioning and development roadmap of the StarWay platform.