Digital Transformation in Process Manufacturing

• Understanding the scope (requirement): Current V/S Expected performance (IT-OT)
• Whether or not the current infrastructure is I-4.0 ready
• How the current data architecture looks like; Sensors, DCS/SCADA, etc.
• Additional sensor Implementation for I 4.0 use cases
• Whether the entity of Data Infra is OPC compliant/evaluating the communication protocols supported
• Identification of IT-OT Data requirement Tech Stack
• Designing ISA 95 compliant IT-OT architecture

Domains: MES, Historians, Industry Data Ops, IT-OT integration, Cloud infrastructure, Unified namespace architecture (UNS), ISA 95, Asset hierarchy

Benchmarking few key performance metrics and Zone to Win- To define Vision, Mission, Goals, Digital Culture, Talent & Skills, Technology Leadership, etc.

• Identifying the pain areas in Process manufacturing
• Blueprinting Dx - Defining the Process Problem Statements – production/ process optimization, Maintenance, Quality, Safety, Sustainability
• Creation of business Case for operational benefits

Operations Areas: Manufacturing planning & schedule, Quality, Overall yield, Zero defects, Energy optimization, OEE, etc.

Financial Analysis - Projected Savings, Benefits report, ROI assessment

• Creating the financial savings report-projected savings per asset/process
• Business case(s) for various tech stack implementations for each identified areas (Process optimization, Maintenance, Quality, Safety, Sustainability)
• Assessment of the required budget
• Estimated savings in the CAPEX, OPEX
• Dx investment planning for 5 years with IRR / NPV / Payback period calculations

For e.g. ROI, Pay-back period, Investment planning, Capex-Opex budget allocation

Develop Priority Matrix depending on the Digital Maturity Assessment, Availability of Assets, Process, People

• Prioritizing the Business problems based on the savings, requirements and challenges
• Identification of Dx Tech Stack(s) for implementation
• Implementation plan and risk mitigation strategy
• Change management
• Industry 4.0 training needs identification
• Cultural transformation roadmap

For e.g. Prioritisation matrix, Dx roadmap, Cultural transformation, Training

PMC for Implementation of Solution(s) – Continuous efforts for e.g., IT-OT integration, Sensor Placements, APC, APM, OTS, Digital Workforce, Others

• Develop Scalability Framework
• Implementation roll-out plan across different locations
• Continuous Improvement - Gap Analysis and Scope of Improvement – KPIs, New / Advanced Use Cases
• Evaluation of Usage of the solution implements
• Post implementation benefit analysis/ ROI calculations
• Identification of gaps in solutions that are implemented and rectification plan
• Identification of Training needs
• New Use Cases definition, New modeling framework, Advanced Analytics, etc.

PMC: Ensuring successful Dx implementation and post-implementation ROI’s

The ISA-95 framework provides standards for integrating enterprise and control systems

• Implement data integrations that link MES, enterprise resource planning (ERP) systems, IIoT platforms, data lakes, and analytics solutions
• Hierarchal model helps to structure and align information from different levels of the corporation -- multiple cells, lines, areas, sites and even enterprises
• Information model - defines the inter-relationships
• Defines the flow of data from different functional models

Domains: ISA 95, Asset hierarchy, MES/ MOM, Industry Data Ops, Unified namespace architecture (UNS)

The data contextualization provides contextualized and standardized data points to the operations team

• Creation ‘Relationships’ using graph database (Knowledge graph)
• Embedding industrial/ domain knowledge to the raw data so that it can become information
• Optimized data, structures, storage and transformation.
• Predefined data models, context models, and combination functions
• Contextual data identifies bottlenecks and opportunities for improvement.

Key Areas: Data enrichment, process / domain contextualization, Knowledge graph

Partner Solutions: Cognite

Provide the tools, process and organizational structures to support the data-focused enterprise

• Creating the financial savings report-projected savings per asset/process
• Business case(s) for various tech stack implementations for each identified areas (Process optimization, Maintenance, Quality, Safety, Sustainability)
• Assessment of the required budget
• Estimated savings in the CAPEX, OPEX
• Dx investment planning for 5 years with IRR / NPV / Payback period calculations

For e.g. ROI, Pay-back period, Investment planning, Capex-Opex budget allocation

Develop Priority Matrix depending on the Digital Maturity Assessment, Availability of Assets, Process, People

• Prioritizing the Business problems based on the savings, requirements and challenges
• Identification of Dx Tech Stack(s) for implementation
• 5-year self-sustainable Dx implementation roadmap (phase-wise) creation
• Implementation plan and risk mitigation strategy
• Change management
• Industry 4.0 training needs identification
• Cultural transformation roadmap

For e.g. Prioritisation matrix, Dx roadmap, Cultural transformation, Training

PMC for Implementation of Solution(s) – Continuous efforts for e.g., IT-OT integration, Sensor Placements, APC, APM, OTS, Digital Workforce, Others

• Develop Scalability Framework
• Implementation roll-out plan across different locations
• Continuous Improvement - Gap Analysis and Scope of Improvement – KPIs, New / Advanced Use Cases
• Evaluation of Usage of the solution implements
• Post implementation benefit analysis/ ROI calculations
• Identification of gaps in solutions that are implemented and rectification plan
• Identification of Training needs
• New Use Cases definition, New modeling framework, Advanced Analytics, etc.

PMC: Ensuring successful Dx implementation and post-implementation ROI’s

• Advanced analytics on timeseries dataset
• Root cause analysis using unsupervised ML techniques
• Open loop recommendations for process improvements
• Tracking KPI deviations and first principle modeling

Techstacks:

• Python-based frameworks
• Supervised and unsupervised modeling
• Flask, Django based web Apps
• Seeq, Cognite, AWS, Azure

• Offers real time prediction for offline lab measurements for quality control
• Translation of low frequency data to high frequency
• Inferentials building using process, mechanical and quality data
• Tracking KPI deviations and first principle modeling

Techstacks:

• Supervised ML techniques
• Boosting and bagging methodologies

• Physics informed ML models for high fidelity predictions
• 1st principles driven ML model for reliable predictions
• Real time optimization of process through neural networks and plant constraints

Techstacks:

• Hugging face APIs
• GPT-3, GPT-3.5, BERT
• Language representation models
• Zero shot model

• % O₂ prediction in flue gas for fired heaters
• Real time Yield prediction in batch processes
• Real time %LOD prediction in batch dryers
• Feed strategy optimization in aluminum smelters
• Channelling prediction in blast furnace operations
• BTP temperature prediction in sinter plant operations
• Real time IBP prediction in crude distillation unit

• Advanced 3D models-based visualization for the telemetry data modeling
• Integrated data management – Timeseries, SAP, MES, ERP, AIMS, and others engineering information
• Analyze the complete history of the processes and equipment
• Knowledge graphs for better information management through metadata modeling

Techstacks:

• Knowledge graphs
• Master data management
• Cognite, AWS, Azure, VidyaTec

• Mimic the plant conditions by connecting all the upstream and downstream equipment
• Improve the decision making through equipment level correlations and causalities
• Enhance the process and equipment reliabilities through sequential models

Techstacks:

• Supervised ML techniques
• Boosting and bagging methodologies
• Seeq, Cognite, AWS, Azure

• Open source based tailored frameworks focused to provide the custom insights for point problems
• Applications focused to accelerate the scalability and adoption
• Built to deliver high value through physics based modeling for real time constraint modeling and optimization

Techstacks:

• Python, scipy, Tensorflow, pytorch
• Supervised and unsupervised based learning
• Fisrt principles based mechanistic modeling

• Solutions leveraging neural networks for image processing and video analytics solving complex industrial problems
• Leveraging artificial neural networks for fault detection through complex learning patterns
• Digitization of offline scanned manufacturing records through OCR
• Analyzing complex microstructures and visual quality inspection

Techstacks:

• Deep learning frameworks – Tensorflow, Pytorch - CNNs, RNNs, GAN
• Generative model, transformers - GPT-4 and others

• Improving the decision making through connected analytics for CDU operations
• Digital twin for complex refinery and petrochemical units
• Plant wide optimization
• Smart Apps for process/energy/utilities optimization for manufacturing companies
• Compressors
• Exchangers
• Steam optimization
• Energy management
• Steam and condensate balance
• Image analytics for safety measures
• Complex material microstructure pattern prediction
• Prediction of external cracks for metal casts

• Advanced process control for complex and dynamics chemical operations for autonomous operations
• Integrated PID controllers for multivariate control strategy to dynamically respond to the process changes
• Minimize process off-spec, enhanced quality control and reduced process variabilities

Techstacks:

• Honeywell Forge, Aveva APC, Aspen DMC

• Offers real time process control through complex analysis of MVs, CVs and DVs
• Integrated non-linear inferentials for real time quality control and constrained objectives / targets
• Accounting the impact of uncontrollable disturbances
• Identifies the opportunity to optimize the processes every minute within the non-linear constrained environment

Techstacks:

• Reinforcement learning
• Non-linear inferentials for quality prediction

• Physics informed RTOs for enhancing the overall performance of the site to meet the cost / objective functions
• Objective oriented model intended to optimize the economic profits
• Continuous production optimization to improve the asset reliability, throughput and energy efficiency

Techstacks:

• Honeywell RTO, Aveva RTO

• Optimize the performance of various processes such as distillation columns, reactors, and compressors
• Reduce energy cost for facilities including utility management
• Optimization of the air-to-fuel ratio in furnaces to improve combustion efficiency, reduce fuel consumption, and minimize emissions
• APC for compressor optimization
• RTO for gas plant optimization
• Stabilization of Distillation columns through APC

Track the carbon footprint under the category of scope 1 and scope 2 emissions.

• Direct emissions from owned or controlled sources
• Indirect emissions from the generation of purchased electricity, steam, heating, and cooling consumed

Domains: Industry Data Ops, Energy management, Carbon abatement

Estimate and predict the emissions of pollutants from critical processes across the manufacturing facilities. Improve the decision making through predictive analytics by optimizing combustion processes and reducing the fuel consumption. Track the emissions – NOx, SO₂, CO, CO₂, PM and others.

Operations Areas: Boiler fuel consumption, condenser and reboiler duty management, HVAC energy management.

Leverage data analytics for GHG analysis and reporting by effectively collecting the emissions, electricity, steam and natural resource consumption across the manufacturing facilities. Benchmark and prioritize the energy consumption for critical equipment for various carbon abatement programs, enabling compliance and sustainability optimization.

For e.g. ROI, Pay-back period, Investment planning, Capex-Opex budget allocation.

• Fuel optimization in boiler operations
• Steam and chilled water consumption in reboilers and condensers
• HVAC energy management 
• Heat integration studies and pinch analysis
• Motors and critical drives energy management

• P&ID mapping and process flow simulators
• Process troubleshooting and debottlenecking
• Process stabilization
• Design modification and process improvement studies
• Heat integration and pinch analysis
• Thermodynamics and process reconciliation

Techstacks:

Aspen Hysys, Honeywell Unisim, Siemens gproms

• Real time plant simulators
• Process benchmarking 
• Process operators training – plant control operations
• Process improvement and benchmarking studies

Techstacks:

Honeywell OTS, Yokogawa OTS, Emerson OTS, Aveva OTS

• Physics informed simulators based on deep learning networks
• Enhancing overall performance of the site to meet the cost / objective functions
• Objective oriented model intended to optimize the economic profits
• Continuous production optimization to improve the asset reliability, throughput and energy efficiency

Techstacks:

• 3D training and simulations using VR
• Enriched remote operations experience
• Immersive assistance in remote maintenance workflows
• Enhanced plant safety
• Enhanced productivity through efficiency and smart operations

• Advanced analytics on timeseries dataset
• Fault classification and categorization
• Analyze the complete history of the processes and equipment
• Provides domain-specific responses and guidance to user prompt
• Helping users navigate complex problems with expert insight

Techstacks:

• Task specific – BART, ALBERT
• Multimodel model
• Advanced Analysis – GPT-4
• Zero shot model
• Few shot LLMs – GPT-3, T5 (Google)

• Offers implicit or explicit active task execution to assist users in achieving their outcomes.
• Streamlining workflows and improving efficiency
• Chatbots for siloed data interpretation and augmentation
• Historical alarm report analysis
• Content generation and language standardization

Techstacks:

• Domain specific models – BioBERT, SciBERT
• GPT-3, GPT-3.5, BERT, Grover
• Language representation model – XLM, Multilingual BERT

• Extraction of meaningful information from siloed and unstructured data
• Vendor management using
• Automation of codes and standards references for asset integrity and fire and safety standards
• Content creation using social media sentiments for improved decision making

Techstacks:

• Hugging face APIs
• GPT-3, GPT-3.5, BERT
• Language representation models
• Zero shot model

• Interpretation of Certificate of Analysis (COA) reports
• Conversational AI for Daily Drilling Reports (DDR)
• Information extraction from Daily Production reports (DPRs)
• Interpretation and analysis of Incident reports for RCA for critical equipment
• Advanced statistical analysis of time series data
• Conversational AI for Batch production using SAP, QMS and Historians
• Sustainability – Energy, Steam/water, waste analysis

• Basic understanding of Time-series / Manufacturing Process Data
• Data Preparation & Pre-processing – Data Cleansing, Conditioning, descriptive statistics
• Process Understanding and how to represent the system from data analytics perspective
• Introduction to Level of Analytics – Descriptive, Predictive, Prescriptive, etc.
• Exploratory Data Analytics - Feature identification, reduction, Covariance methods, basic formula techniques
• Basic introduction to ML libraries & statistical templates and its applications
• Use-cases examples – How Time-series data analytics is used

User Profile: Process engineers, Technologists, Plant operations team, Maintenance & Reliability Enggs., IT teams, etc

Duration: Typically, 1 week. Depends on the Audience Size.

• Introduction to building formulas and statistical/ ML python scripts for advanced analytics (for e.g, Root Cause Analysis, Critical Parameter Identification, Soft Sensors, etc.)
• Advanced Data modeling techniques (using Python, Jupyter Notebook, etc.) on your data and results
• Model construction, application / validation, Visualization and implementation – Right Methods (Use case-specific)
• Introduction to commercial analytics tools
• User Profile / Audience: Users (process domain experts) who have basic coding background (Python) and understanding on various Statistical/ ML algorithms to create Models and apply it for analytics
• Prerequisite: Python expert in ML and its application
• Duration: Typically, 1 week. Depends on the Audience Size.

• Python-based template creation and widget building
• Soft-sensor building (using Physics-based, First-principle Models)
• Advanced reports and Dashboarding
• ML Model(s) deployment methods
• Best Practices / Guide & Methods for Data Analytics
• User Profile / Audience: SMEs / Process domain experts with Mathematical Modeling, Process Modeling, Python/ML expertise
• Duration: Typically, 1 week. Depends on the Audience Size.

• Tridiagonal will provide analytics engineering support to identify and execute the use cases on your data with your users. We will work on the identified three (3-4) Use cases. 
• The prerequisites in terms of the data requirements (signals/tags/volume, etc.), process details, data modeling & analytics objectives will be finalized. The use case can be worked on the Off-line / On-line data, provided the customer give an access to the data /analytics environment. We will submit the analytics outputs and ROI reports to the teams
• Duration: Typically,4-6 weeks. Depends on the number of Use Cases