By Molly Pace, Cement Economics and Implementation Engineer, Imubit
The cement industry does not lack acronyms or challenges. Three abbreviations dominated the hallway conversations at the 2025 IEEE-IAS/PCA Cement Industry Conference (the PCA is now the ACA): AI for artificial intelligence, AFR for alternative fuels and raw materials, and 1L for the ongoing shift to Portland limestone cement.
The conference floor showed an industry that wants to modernize and is honest about what still stands between the goal and plant execution. Alongside the new tools on display, many attendees voiced frustration with the pace of plant modernization and of 1L adoption across North America. These were the main takeaways.
Net zero runs on several parallel paths
The industry's commitment to net zero by 2050 framed nearly every session. Producers recognize that incremental changes will not close the emissions gap fast enough. Speakers argued for integrated programs that tie kiln upgrades, alternative fuels and data-driven optimization into one investment case that a finance team can support.
AFR helps, once its constraints are acknowledged
AFR adoption is advancing, but plant managers described a familiar set of internal and external hurdles:
- Equipment constraints, and capital allocation for feed-system retrofits
- Competition with landfills, which remain the cheaper disposal route where there are no incentives
- Legislation that keeps changing and varies by state, province and country
- Variable feedstock quality, and uncertainty about long-term fuel availability
- Process control complexity that puts product consistency at risk
The consensus among attendees was that clearing these hurdles takes a mix of technology and cross-functional governance. Procurement, environmental and site operations teams need one shared view of emissions and economics.
Portland limestone cement now turns on logistics
The transition to Portland limestone cement (PLC, or 1L) continues to gain momentum on the strength of its lower carbon intensity. Technical debate about strength development is fading, and the discussion has moved to the next hurdle, which is logistics. Fine limestone sourcing, silo space and market acceptance are challenges every producer faces. Speakers advocated coordinated rollouts that keep production, distribution and specification updates in step.
AI in cement, from predictive maintenance to closed-loop control
AI is already at work in cement plants through predictive maintenance. These tools analyze sensor data, equipment performance trends and failure history to forecast when parts will need replacing and where costly breakdowns are likely. That has become important for plant availability, downtime and maintenance budgets.
Predictive maintenance answers when equipment is likely to fail. A broader question is how all the variables in your specific process interact. Kiln and calciner behavior is complex, dynamic and nonlinear. Once those relationships are understood, the same plant data can support operating decisions as well as maintenance decisions.
The organizational side
Presenters returned to one theme: the industry's challenges are as much organizational as technical. Winning budget for decarbonization takes a case that ties cost savings to sustainability benefits. Talent pipelines need to develop "translator" roles that connect process engineering and data science. And in a sector where kiln uptime is the priority, a change initiative succeeds when it supports stable production rather than disrupting it.
What this means at the kiln
Several of these themes meet at the kiln and calciner, where fuel substitution, product quality and emissions have to be balanced continuously.
Imubit's approach starts with the plant's own data, from all operating modes, all feedstocks and all AFR substitution levels. Deep Learning Process Control® (DLPC®) learns a nonlinear dynamic model of the pyroprocess from that history, guided by first principles, and pushes toward the objective the plant defines, whether that is margin, emissions or a balance of the two. Procurement, environmental and operations teams, which have often worked from separate views of the process, then work from one model of how the plant behaves.
The models detect and learn nonlinear relationships between variables that are hard to see in daily operation, and they pick up shifts in the data that would otherwise go unnoticed. Figure 1 shows gain histograms for three controlled variables against alternative fuel rate. For calciner gas inlet temperature, the gain flips from negative (red) to positive (blue) across the operating range. A linear model with a single fixed gain cannot represent that change of sign.
Figure 1. Gain histograms from a preheater and calciner model, showing the response of calciner gas inlet temperature, preheater outlet O2 and preheater gas outlet temperature to alternative fuel rate. The calciner inlet temperature gain flips from negative (red) to positive (blue).
This matters for AFR, where the interplay of combustion and material variability is difficult to quantify. A model trained on plant data does more than reproduce what the team already knows. It can show relationships the team has not been able to measure.
In closed loop, DLPC® writes setpoints minute by minute on top of the plant's existing APC and DCS, with operators able to take over at any time. For more on fuel substitution in the kiln and calciner, see alternative fuels in cement kilns and AI-enabled combustion control, or read about Closed-Loop Physical AI for cement.
