How Semiconductor Manufacturing Shaped OPTIFICIENT AI Solutions

1. Industrial Analytics has developed industrial AI solutions since 2017 and became part of Infineon in 2022. How did the demanding environment of semiconductor manufacturing shape OPTIFICIENT?
a. Lisa Erdmann, CEO of Industrial Analytics: “Semiconductor manufacturing set a very high bar for OPTIFICIENT™. Optimization must never compromise process stability, air quality, availability, or product quality. In a fab, heating, cooling, humidity, pressure, and air treatment are closely linked to production. An AI system therefore cannot simply identify a statistical correlation and change a setpoint. It must understand the physical system, operate within defined limits, and make every action traceable. That is why OPTIFICIENT combines machine learning with thermodynamic HVAC models and engineering rules. It continuously identifies more efficient operating conditions and adjusts setpoints within an approved window of operation. This allows customers to improve the efficiency of existing assets while maintaining all operational requirements.”

2. What advantage does your semiconductor experience provide when optimizing commercial, medical, and industrial properties?
a. Lisa Erdmann: “The operating targets differ, but the engineering challenge is similar: existing assets, changing loads, interacting control loops, fragmented data, and limits that must not be exceeded. Semiconductor sites taught us to address these factors under particularly demanding conditions. That experience transfers well to hospitals and data centers, where availability and environmental limits are also critical, and to offices and shopping centers, where comfort and indoor-air requirements define the boundaries. OPTIFICIENT adapts the optimization objective to each application and works with the installed infrastructure. It can optimize temperature and humidity setpoints, equipment sequencing, air-treatment strategies, airflow setpoints, and load schedules while respecting process, comfort, and operational constraints.“ Thus, customers benefit from rock solid, reliable software that has been tested and refined under demanding conditions.”

3. What evidence do you have from operating OPTIFICIENT in Infineon facilities and other demanding environments?
a. Lisa Erdmann: “OPTIFICIENT has been validated in real production and building operations, including Infineon facilities and manufacturing sites. Two completed projects illustrate the measurable value. In a semiconductor cooling-water application, for example, with four chillers, four cooling towers, and eight pumps, optimized chilled-water and condenser-water setpoints delivered annual savings of 402 MWh and energy-cost savings in the five-figure euro range. In a 60,000 m² cleanroom air-treatment application, optimization of the thermodynamic path across six multi-section systems delivered annual savings of 7,900 MWh and energy-cost savings in the mid six-figure euro range. All this while maintaining strict air-quality requirements.

In a separate, ongoing learning project with energy efficiency and technical facility-management partner SPIE Germany Switzerland Austria and Infineon subsidiary SILTECTRA™, we are applying the same repeatable method to cleanroom air handling: assess the data and available degrees of freedom, define safety limits, model the HVAC system, apply optimization algorithms, deploy the software, and verify the savings. It is also helping us develop a partner model that can be replicated across further sites.”

4. How does your hybrid-AI approach make optimization transparent and trustworthy for facility managers?
a. Lisa Erdmann: “Hybrid AI is central to trust because it turns engineering expertise into scalable, transparent operations. A purely statistical model may identify a pattern that looks promising in historical data but is unsuitable under a different load, weather condition, or operating mode. OPTIFICIENT therefore evaluates data-driven opportunities against thermodynamic HVAC behavior, equipment characteristics, engineering rules, and site-specific boundaries.

Operators can actually see which variables support an adjustment, what effect is expected, and which limits remain protected. At the start of a project, we agree constraints, approval thresholds, alarm limits, and rollback procedures with the customer. In practical terms, we use internal AI tooling to do much of the heavy lifting in pre-processing large volumes of documentation and operational data. Our engineers then review and validate the models and resulting model adjustments. They work closely with the facility team with AI helping to accelerate the analysis process rather than replacing engineering judgment. The customer determines and retains control over the permitted level of autonomy. The result is not a black box, but an optimization approach whose recommendations, and automated actions can be explained in technical terms.”

5. How does OPTIFICIENT move beyond monitoring and recommendations to active optimization, and how does this change the work of facility-management teams?
a. Lisa Erdmann: “A conventional building management system is essential for monitoring, control, schedules, and alarms, but it does not normally calculate the most efficient operating point across an interconnected HVAC system. OPTIFICIENT closes that gap by continuously reading operational data, modelling system behavior, identifying more efficient operating conditions, and transferring validated setpoint adjustments through the existing control environment. These can include chilled-water and condenser-water temperature setpoints, airflow and humidity setpoints, and load distribution across multiple assets. For facility teams, the aim is not to remove engineering responsibility. It is to automate repetitive analysis and continuous fine-tuning, identify emerging system issues, and allow operators to focus on exceptions, maintenance, and decisions that require site knowledge. Operators retain transparency and can override optimized settings when necessary; the system then makes the effect of that intervention visible.”

6. What safeguards protect equipment and operational limits when OPTIFICIENT writes setpoints to existing building-automation systems?
a. Lisa Erdmann: “Safety starts before any setpoint is written. During assessment and commissioning, we verify data quality, assess the existing control structure, identify the adjustable degrees of freedom, and define the operating conditions with the customer. This includes room or process conditions, equipment limits, minimum ventilation, redundancy rules, alarm thresholds, and permitted rates of change. Every setpoint write is validated against those rules and introduced in controlled, reversible stages.

The existing building management system, distributed control system, or controller retains its native protection and alarm functions. Service-level watchdogs monitor whether the optimization service is active and delivering. The response to a lost data connection, communication failure, or invalid operating state is defined and tested for each site rather than presented as a universal fail-safe. Depending on customer requirements, the service can be deployed in the cloud or on premises. In one of our current assignments, it runs on a virtual machine at the site so operational data does not need to leave the premises.”

7. Before deployment, you conduct an AI Optimization Check. What common data or system issues do you uncover, and how do you resolve them?
a. Lisa Erdmann: “The AI Optimization Check is designed to identify issues before automation begins. Typical findings include missing or implausible sensor values, inconsistent names and units, gaps in trend data, undocumented control logic, drifting sensors, and differences between intended and actual plant behavior. We treat these as engineering gates, not as problems for the AI to ignore. The team validates data quality and the condition of the HVAC system, then defines which optimization measures can be supported reliably. Where documentation and operating data do not align, we correct mappings, validate sensors or meters, and narrow the scope until the evidence is sufficient. The resulting report covers data quality, system condition, simulated savings potential based on historical operation, and integration recommendations. In this process knowledge graphs help us structure customer documentation and HVAC relationships more quickly, but engineering validation remains essential.”

8. What determines whether a facility achieves savings at the upper or lower end of the observed 5 to 30 percent range?
a. Lisa Erdmann: “That very much depends on the starting point with regards to existing efficiency and the number of interacting variables that can be adjusted dynamically. Building type, baseline operating conditions, data availability and quality, and existing control strategies all influence the outcome. A site with static operating strategies, conservative safety margins, or poorly coordinated controls will generally offer more potential for improvement than a well-tuned facility with limited operational flexibility. That is why savings can range from 5 to 30 percent. This is evident from completed projects, we will not guarantee outcome prior to thoroughly looking into the project. For this, we assess each facility individually, establish a normalized baseline, and verify savings against its actual operating conditions. AI-driven optimization can deliver measurable value while maintaining comfort, process requirements, and reliability.”

9. Why offer OPTIFICIENT as a white-label SaaS solution through facility-management and building-automation providers rather than compete with them directly?
a. Lisa Erdmann: “Facility-management and building-automation providers already have established customer relationships, understand the installed systems, and carry responsibility for service levels across large portfolios. Rather than compete with that expertise, we believe a partnership model creates greater value. We provide a scalable optimization capability without requiring partners to build specialist AI, thermodynamics, and data-science teams from scratch. Because OPTIFICIENT integrates with existing building-automation environments, providers can retain their customer interface, hardware ecosystem, and operational processes while adding measurable HVAC optimization under their own brand. Our collaboration with SPIE Germany Switzerland Austria and SILTECTRA shows how this can work in practice. Partners contribute customer access, technical operating expertise, and portfolio reach; we provide the optimization technology, engineering models, and implementation support.”

10. Where is OPTIFICIENT already being used, and how do you plan to scale internationally?
a. Lisa Erdmann: “OPTIFICIENT is designed for global deployment. As a vendor-independent software add-on, it can run in cloud or on-premises environments and connect through widely used protocols such as BACnet, Modbus, OPC UA, and MQTT. This allows customers to optimize heterogeneously installed HVAC assets without an immediate hardware overhaul and supports different data-governance requirements across international portfolios. Industrial Analytics has project experience with customers and sites in several markets, including Singapore, Malaysia, Italy, and Austria. The partner model is designed for replication across multiple customer sites. Our immediate priority is to scale with facility-management and building-automation partners in markets where customers operate large, energy-intensive portfolios and have accessible building or industrial control data.”



Industrial Analytics CEO Lisa Erdmann

(1) Thermodynamic path optimization focuses on selecting the most efficient sequence of air treatment steps within an Air Handling Unit (AHU) to achieve the required supply air conditions. Instead of heating, cooling, humidifying, or dehumidifying more than necessary, the system optimizes each step, so air reaches the target temperature, humidity, and cleanliness for semiconductor manufacturing with the lowest possible energy use while maintaining strict process requirements.

Interview conducted by Maria Brueva, Induportals editor.

www.industrial-analytics.io