Skip to content

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

LayerSystem TypeTime ScaleCore Tasks
L4 Enterprise LayerERP/MESDay/Week/MonthProduction planning, material management, quality traceability
L3 Optimization LayerAPC/RTOMinute/HourMultivariable coordination, economic optimization, constraint pushing
L2 Control LayerDCS/PLCMillisecond/SecondSingle-loop control, logical interlock, safety protection
L1 Field LayerSensors/ActuatorsMillisecondSignal acquisition, action execution

1.3 Positioning of Data Insight Platform

Analysis:

  1. 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).
  2. 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.
  3. OPC Server / DCS / PLC (Basic Control & Communication Layer): Receives the setpoints from StarWayODI and executes physical actions such as valve openings and pump speeds.
  4. 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:

ScenarioControl TaskCommon Brands
Packaging MachinerySequence control, position controlSiemens, Mitsubishi, Omron
Production LinesStart-stop control, speed coordinationRockwell, Schneider
Water TreatmentPump-valve interlock, liquid level controlSiemens, ABB
Building AutomationAir conditioning control, lighting controlHoneywell, 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:

IndustryApplication CharacteristicsMainstream DCS Brands
PetrochemicalHigh temperature and pressure, continuous productionHoneywell, Emerson, Yokogawa
PowerLarge unit coordinated controlHollysys, Guodian Zhishen
PharmaceuticalBatch control, GMP complianceSiemens, Rockwell
MetallurgyLarge-scale, high reliabilityABB, Siemens

DCS vs PLC Comparison:

Comparison ItemDCSPLC
System ArchitectureDistributed, networkedCentralized or distributed
Control ScaleLarge-scale (thousands of points)Small and medium-scale (hundreds of points)
Control TypeMainly continuous process controlMainly discrete logic control
Engineering CostHigh (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)

  1. Predictive Model: Establish input-output dynamic model based on historical data
  2. Rolling Optimization: Solve open-loop optimization problem in each control cycle
  3. Feedback Correction: Use actual measured values to correct model predictions
  4. Implementation of First Step: Only implement the first control action of the optimization results

APC Application Effects:

Application ScenarioTypical BenefitsImplementation Cycle
Distillation Column ControlProduct purity increased by 5-10%, energy consumption reduced by 3-5%2-3 months
Reactor ControlConversion rate increased by 2-5%, selectivity improved3-4 months
Furnace ControlThermal efficiency increased by 2-4%, emissions reduced2-3 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

  1. Data Preparation (2-4 weeks): Collect historical data, data quality inspection
  2. Exploratory Analysis (1-2 weeks): PCA analysis, identify abnormal batches
  3. Modeling Analysis (2-3 weeks): PLS modeling, model verification
  4. VIP Analysis (1-2 weeks): Identify key variables, develop optimization recommendations
  5. Verification and Implementation (4-8 weeks): Experimental verification, develop SOP, training implementation

3.3 Model Acceptance Criteria

IndicatorMinimum RequirementGoodExcellent
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!

DimensionStarWayDIAPC
PositioningOffline analysis toolOnline control system
DataHistorical dataReal-time data
ModelStatic model (PLS)Dynamic model (MPC)
OutputAnalysis reports, optimization recommendationsReal-time control commands
ExecutionManual executionAutomatic execution

Collaboration Process:

  1. StarWayDI analyzes historical data, identifies key variables
  2. StarWayDI establishes static PLS model, evaluates feasibility
  3. Transfer model/variable information to APC implementation team
  4. APC engineers establish dynamic MPC model, connect DCS control
  5. 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:

FeatureStarWayDIStarWayODI
DataHistorical dataReal-time data access
ModelStatic modelDynamic model update
ExecutionManual execution of optimization recommendationsAutomatic issuance of optimization commands
OperationRegular analysisContinuous 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:

LayerSystemOperation CycleCore Tasks
RTOReal-time optimization layer15-60 minutesEconomic benefit optimization, setpoint optimization
APCAdvanced control layer1-5 minutesTrack RTO setpoints, multivariable coordination
DCSBasic control layerMillisecond-second levelPID 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

PhaseKey Success FactorsRisk Points
Offline SetupData quality, engineer trainingData missing, resistance from field staff
In-depth ApplicationBusiness value verification, management supportAnalysis results fail to guide actual production
Online Control UpgradeModel accuracy, OPC communication stabilityDynamic model mismatch, underlying PID loop anomalies
Closed-loop & Self-learningSystem adaptive capability, network securitySystem failure under extreme conditions, data security risks

7.3 Return on Investment Analysis

Investment PhaseMain Investment ContentExpected Core BenefitsPayback Period
StarWayDI (Offline Phase)Software license, offline data implementation servicesEnhance process cognition, discover optimization potential, 2-5% process optimization6-12 months
StarWayODI (Online Phase)Software license, IoT/TSDB hardware & interface development, control engineering servicesAchieve closed-loop automatic control, 3-8% energy consumption reduction, yield improvement12-18 months
Plant-wide RolloutMulti-unit expansion licenses, continuous maintenanceComprehensive benefit improvement, moving towards dark factories / unmanned operation18-24 months

Appendix: Terminology Quick Reference

TermEnglish Full NameChinese Explanation
PLCProgrammable Logic Controller可编程逻辑控制器
DCSDistributed Control System分布式控制系统
APCAdvanced Process Control先进过程控制
MPCModel Predictive Control模型预测控制
RTOReal-Time Optimization实时优化
OOOffline Optimization离线优化
MESManufacturing Execution System制造执行系统
ERPEnterprise Resource Planning企业资源计划
OPCOLE for Process Control过程控制OLE
SCADASupervisory 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.

Let data speak, make decisions simpler.