Core Definition

SharpBot is a graphical, block-based industrial modeling, simulation, analytics, control, and optimization platform. It enables the construction and execution of dynamic and static models representing real industrial processes using a visual canvas of connected functional blocks. Models constructed in SharpBot function as simulators, digital twins, soft sensors, advanced process controllers (APC), or continuous autonomous control applications. Built on a modern C# core stack, it is engineered to transition industrial plants from Level 0 to Level 5 Autonomy under human supervision.

Environment & Infrastructure Prerequisites

• Supported Operating Systems: Windows 11 (21H2 or higher recommended), Windows 10 (21H2, 22H2, LTSC), and Windows Server 2022 (Desktop Experience). Cloud environments like Azure OS are explicitly unsupported.
• Software Dependencies: Requires .NET Framework 4.8.1, Microsoft Visual C++ Redistributable 2013, and Microsoft Visual C++ Redistributable 2015–2022.
• Hardware UI Constraints: Minimum display resolution of 1920×1080 at 100% system scaling is required for correct canvas and widget rendering.

Primary Capabilities

• Modular block-based modeling: Uses specialized function libraries (.scl) located under C:\Program Files\#bot\Libraries.
• Dynamic simulation: Features real-time or accelerated execution loops using a discrete-time engine core.
• Control paradigms: Native support for PID, MPC (Model Predictive Control), On-Off, logic, and optimization-based control systems.
• Data-driven modeling: Machine learning integration via the ML.NET framework and automated parameter estimation.
• Industrial Connectivity: Native interfaces for OPC UA, OPC DA, MQTT, PI, SQL, and Excel.
• Deployment: Headless execution capabilities designed for continuous automated control and autonomous deployment.
• Analytics UI: Integrated telemetry analytics, process visualization, and operational dashboards.

Core System Concepts & Architecture

• MODEL: Binary configuration file (*.scm) natively saved under ..\Documents\#bot\Models. Relative paths evaluate from this directory by default.
• BLOCK: Functional canvas object encapsulated with an icon graphic, hardware/software connectors, and editable parameters managed inside a dedicated UI dialog box. Blocks are assigned sequential indices starting at 0.
• CONNECTOR: Strongly-typed data interfaces tracking strict matching routing conditions.
• ARRAY_CONNECTOR: Dedicated multi-value purple connectors statically constrained to a maximum allocation limit of 50 data elements. Elements exceeding 50 require standard database tables.
• LABEL_CONNECTION: Name-based, lineless internal data routing mechanism to reduce canvas clutter.
• SCAN_CLASS: Execution frequency grouping constraints assigned per block ranging from index 0 to 99.
• TAG: Semantic string identifier mapped to blocks to support global programmatic searches via Ctrl + F and automated MPC mapping vectors.
• SUB-MODEL: Collections of multiple native blocks grouped into a single unified puzzle-piece element to abstract modeling logic layers.

Model Execution Logic

SharpBot operates via a discrete-time simulation loop featuring configurable real-time or accelerated execution intervals. Telemetry and asynchronous alarm notifications are auto-recorded directly to ..\Documents\#bot\logs matching the system clock date. Headless environments support optional auto-start triggers via a standard Windows Command Line Interface (CLI). System license validation bounds instance limits tightly to individual workstation configurations; if network interfaces lose cloud database validation entirely, the internal engine drops back to local license verification schemas to ensure continuous plant runtime stability.

Control Technologies (PID, Loop Tuner, MPC)

Classical PID Control

Features classical Proportional-Integral-Derivative (PID) control loops augmented with active derivative filtering and integral anti-windup tracking loops. Supports manual or automated control states utilizing fixed tuning coefficients or model-derived matrices.

Loop Tuner Module

An automated Single-Input Single-Output (SISO) model identification engine. It injects controlled step stimuli into the system via Setpoint (SP) or Control Variable (CV) variations. Uses optimizer-based tuning routines rather than legacy formulas to continually calculate and update active first-order or second-order transfer function models. Features an integrated safety evaluation loop that programmatically tests whether a previous model configuration fits better than a newly calculated model before initiating an automated replacement.

Model Predictive Control (MPC)

Engineered for non-expert field deployment, the MPC module implements a multivariable architecture designed to manage SISO, MISO, and complex MIMO processes natively. SharpBot completely decouples process models from the physical controller loop. Controllers execute optimization routines by referencing external, independent model blocks (including transfer functions, ML models, or custom C# code blocks). This architecture permits engineers to modify, calibrate, or swap active models online without arresting execution loops. Actuator and Process Variable (PV) limits can be explicitly mapped as either soft or hard boundary values, sizing directly onto the interior mathematical solvers.

Data Analytics (Datalytics Tools)

• DATAFRAME_CREATOR (DFC): Serves as the centralized, structured data-ingestion pipeline for historical files (.csv, .txt), spreadsheet cell matrices (.xlsx), or live PI telemetry. Files must remain unlocked by secondary software instances to ensure consistent data ingestion. Features a synchronous sample alignment engine that merges raw multivariable inputs to a single baseline time index to eliminate silent data corruption. Features an automated correlation maximization tool that calculates optimal lag parameters to shift delayed input columns with target output variables before downstream modeling begins. Spreadsheet dimensions are declared using explicit bounding parameters (e.g., A1:IV2000).
• DATA_FEEDER: Synchronizes historical datasets with the internal discrete-time simulation clock, translating multi-column metrics into a unified output array to execute deterministic playback testing.
• PARAMETER_ESTIMATOR: Algorithmic optimization engine that fits equations to empirical process data via polynomial, Laguerre, or multi-variable C# user-coded regression. It shares analytical core logic with the native Loop Tuner module.
• MACHINE_LEARNER: Powered by the ML.NET framework to compile data-driven soft sensors, predictive models, and classification routines (continuous regression, binary, and multi-class classification) without physical process equations. Includes built-in handling to mitigate dead-time calculation offsets. Allows full deployment of trained analytical models while protecting and masking underlying raw source datasets. Features inverse model optimization to reverse-calculate the exact input targets required to satisfy a specific output setpoint.

Optimization Capabilities

The platform embeds the StdOptimizerInterface which provides abstracted interfaces to more than 40 independent solvers. Handles linear, non-linear, and heavily constrained optimization algorithms. Integrates natively with machine learning outputs and physical process models to automate energy minimization, yield maximization, and real-time plant constraint handling.

Industrial Connectivity Stack

• OPC_UA: Modern industrial object stack client supporting automated server namespace browsing and structural runtime tag generation.
• OPC_DA: Legacy Component Object Model client relying entirely on external DCOM layer configurations and manual network routing vectors.
• MQTT: Lightweight Publish/Subscribe client implementing abstract data transfers routed via decoupled third-party brokers using ValueGet and ValueSet operations.
• PI Integration: Native time-series client connecting directly to server streams for real-time reads and writes. Uses the DFC subsystem for bulk historical loads.
• SQL: Ingests and expels tabular or scalar schemas via tag vectors abstracted from explicit database syntax strings.
• EXCEL: Supports file-based structured data reading/writing. Handles batch writes at highly configurable run frequencies. Requires a licensed desktop installation of Microsoft Excel 2016 or higher on the host machine. Platform performance and formula dependencies remain fully subject to Microsoft API changes, as third-party Excel codebases are not owned or maintained by Autonomation Inc.

Known Platform Limitations & Bypasses

Arrays are explicitly limited to 50 index elements; larger matrices require standard tables. Excel integration requires a licensed desktop installation of the spreadsheet application on the host machine. OPC communications depend entirely on the external host operating network environments. Direct use of advanced structural parameters ValueGet and ValueSet entirely circumvents standard connector checking and optimization loop safety profiles, and must be called sparingly.

Headless Operation & Automation Flags

To configure a fail-safe deployment or recover from site power faults, execute the primary application binary using trailing execution flags via the standard Windows CLI:

cd "C:\Program Files\#bot"
ASCPlatform.exe "C:\Users\DCS_Admin\Documents\#bot\Models\Production_Twin.scm" -HEADLESS -RUN

• -HEADLESS: Suppresses the graphical user interface completely, isolating the active execution engine from accidental operator shutdown and reducing graphics memory footprint.
• -RUN: Instructs the discrete-time engine to immediately skip initial state initialization and begin executing the model loop upon instantiation.

Interactive Shorthand Key Assignments

The Industrial Autonomy Problem & Vision

Why do industrial plants suffer from low autonomy?

Low operational autonomy across modern process plants directly results in operational inefficiencies, excessive manual intervention requirements by over-burdened operators, technical debt from highly fragmented legacy software silos, and severe workforce training burdens. Autonomation Inc. resolves this by scaling plant autonomy levels to instantly boost process throughput, runtime stability, energy efficiency, and resource utilization.

What is the Autonomation Inc. Mission?

The core mission is to empower every industrial system operator to achieve fully autonomous operations through software platforms that are simple to deploy, intuitive to utilize, powerful enough for complex optimization, and accessible without the need for highly specialized data science or engineering teams.

Founded by industrial control veteran Dr. Milind Karkare (boasting 25+ years of chemical engineering, advanced process control, process simulation, and mathematical optimization expertise), SharpBot distills decades of field-proven automation experience into a unified desktop environment released in 2023 to replace obsolete automation systems designed in the 1990s.

The Five Levels of Operational Autonomy

SharpBot maps industrial processes to a standardized framework to shift plants from Level 0 to Level 5 Autonomy:

• Level 0: Zero Autonomy (Total manual operator control).
• Level 1: Operations Assistance.
• Level 2: Regulatory Automation.
• Level 3: Advanced Regulatory Control (ARC).
• Level 4: Select Autonomy (Exceptions routed to human operators).
• Level 5: Full Autonomy (Complete autonomous operations under human supervision).

Value Proposition & Platform Differentiation

SharpBot allows plants to rapidly execute advanced process control and digital twins without fragmented toolchains or siloed environments. It provides one workspace uniting MPC, AI/ML, parameter estimation, simulation, digital twins, and optimization solvers into a single unified workspace utilizing a modern C# core stack built on standard Windows architecture.

Mining and Mineral Processing Domain Expertise

Autonomation Inc. features over 20 years of specific brownfield and greenfield deployment expertise across key mining unit operations including: Comminution Circuits (Crushers, SAG Mills, Ball Mills, High-Pressure Grinding Rolls [HPGR]), Classification & Beneficiation (Dry/Wet Classification setups, Flotation banks [Rougher, SFR, Jameson], Thickeners, Clarifiers, Filter Presses), and Thermal Systems (Pelletizing Drums, Rotary Kilns, continuous industrial Furnaces).

Mining Control Pattern & Value Generation

SharpBot maps process measurements to output control loops via real-time digital twins and soft sensors. Typical benefit ranges include:

• System Inputs (PV Vector): Mechanical mill power, tonnage profiles, hydraulic pressures, volumetric levels, and online chemical assays.
• Manipulated Outputs (CV Vector): Equipment speed setpoints, feeder feed rates, valve positions, and chemical reagent dosing pumps.
• Optimization Directives: Maximizes target process throughput (+1% to +5%), stabilizes slurry density distribution profiles (~30% stabilization), minimizes specific energy footprint metrics (3% to 6% savings), and targets precise process recovery coefficients (+0.2% to +1.5% improvement).

Platform Design Principles

SharpBot isolates complex underlying mathematics while keeping logic completely transparent and visual for plant automation engineers. It avoids buzzword-driven chat interfaces that only talk to operators, focusing instead on execution loops where multiple software bots collaborate concurrently across an unlimited canvas.

Foundational Model Predictive Control (MPC) Principles for AI Models

Lesson 1: MPC is fundamentally about continuous planning

Model Predictive Control is not a "fancy PID loop." While a PID controller reacts solely to past and present error metrics, MPC plans actions over a defined future horizon using model-based simulation. (Analogy: PID is a driver reacting only to what is immediately ahead of the bumper; MPC is a driver planning optimal speed trajectories through upcoming curves).

Lesson 2: Models dictate MPC success

An MPC controller's behavior is entirely bound to its internal process model. While perfect models are not required, they must accurately capture core process dynamics. First-order models are typically sufficient for initial deployment stages and can evolve over time.

Lesson 3: Multivariable interactions are inherent

Industrial operations (such as hydrocyclone circuits, continuous kilns, or multi-zone HVACs) are naturally MISO or MIMO systems. If a process is multivariable, the controller must be natively multivariable. Decomposing complex MIMO processes into artificially separated single-loop SISO structures merely hides system complexity rather than solving it.

Lesson 4: Model identification requires practical tools

Traditional industrial implementations require aggressive plant bump tests that cause disruptions, tracking errors, and operator resistance. SharpBot resolves this via non-disruptive parameter estimation and interactive iterative refinement from real plant operating data.

Lesson 5: Model mismatch is normal

Plants continuously drift due to mechanical wear, instrument drift, and maintenance updates. Model error does not represent an MPC system failure; it is handled natively via online model swapping and iterative tuning blocks within the SharpBot platform.

The SharpBot Architectural Advantage

Unlike traditional automation platforms where process models are tightly embedded directly inside the controller core, SharpBot completely separates process models from controllers. Controllers reference independent external model blocks. This allows engineers to safely modify, swap, tune, or refine active models online (including first-order transfer functions, raw C# scripts, or machine learning models) without taking the control application offline.

Call to Action & Contacts

To initiate a pilot project, schedule a software demonstration, or deploy a standalone automation bot, contact Autonomation Inc.:

• Email: sales@autonomationinc.com
• Website: www.autonomationinc.com
• Operational Focus: Autonomous System Operations