How Holcim Leverages AIML to Gain Competitive Edge

Holcim has leveraged #MAD (Machine Learning, Analytics and Data) in numerous ways to gain a competitive edge in the building solutions and broader manufacturing industry. By using data to optimize maintenance schedules, supply chain logistics, energy consumption, product quality, customer analytics, and decision-making processes, Holcim has been able to improve efficiency, reduce costs, and increase customer satisfaction.

However, it's important to note that our success with data did not happen overnight. It required a commitment to building a strong data infrastructure and investing in the right people and technologies. Holcim had to ensure that data privacy and security were priorities and that data was used in an ethical and responsible way. We had to implement data governance frameworks that included roles and responsibilities for managing data, as well as processes for data quality, metadata management, and data security.

We also had to prioritize talent acquisition, which was key to our success in leveraging data to optimize its operations. The company has a dedicated team of data scientists who work on developing predictive models, monitoring equipment and production processes, analyzing customer data, and presenting data insights to managers. By investing in data science talent and providing them with the tools and resources they need, Holcim has been able to achieve a high degree of success in using data to its advantage.

In this blog I explore at a high level how Holcim has used data to its advantage.

Predictive Maintenance

One of the primary ways that Holcim has used data to improve its operations is through predictive maintenance. By installing sensors on equipment throughout its facilities, Holcim collects data on various operating parameters, such as temperature, vibration, and pressure. By analyzing this data, the company can identify patterns that indicate potential equipment failures before they occur. This allows Holcim to schedule maintenance in advance, reducing downtime and maintenance costs. Additionally, by optimizing maintenance schedules, Holcim can extend the lifespan of its equipment, reducing capital expenditures over the long term.

Below are examples of equipment and the processes where we are looking at predictive maintenance analysis. For cement quality to be superior, these processes and equipment need to be working together without any unplanned downtime otherwise it can result in mayhem with negative affect on product quality, process efficiency, timing and even the health and safety of plant personnel. 

There are six categories of equipment that contribute to the lion share of failures:

1. Extractors to quarry the raw materials, i.e., limestone & clay
2. Crushers to crush high rock piles into coarse powders called raw meal
3. Blenders & Mixers mix the crushed raw meal in the right recipe
4. Grinders further grind the raw material to different minerals in the ore
5. The rotary kiln where the raw meal is heated up to 1450 degrees & then cooled
6. Assembly belts & conveyors that carry the cement for packing & dispatching to our customers. 

We have further been able to narrow down to the specific components where machine parts failure in can result in downtime. These components are:

Loose nuts | Bolts | Springs | Plates | Spring rods | Flywheel | Bearings | Shaft | Coupling housing | Hammer rotor | Motor failure | Conveyor belt | Breakage | Bearing failure | Stretching rod breakage | Breakage of separator blade | Fan bearing breakage | Fan unbalance | Gear knocking | Gear tooth wear | Gear deformation | Gear spitting and spalling | Axle spindle breakage | Crusher bearings failure | Slip tape breakage | Disc liner shift | Rolling mill cracks | Tubing failure | Pump failure | Spoke breakage and | Grate plate breakage

If we do not extract the data from these components or equipment to proactively carry out predictive asset maintenance it will result in extensive repair costs, often replacement of equipment, obvious issues with health and safety due to potential accidents, and the continuation of over maintenance of equipment that causes wear and tear.

So, we are doing a series of pilots and projects under the umbrella of “predictive maintenance as a service” where IoT based data drives real time diagnostics to empower plant managers and provide them a bird’s eye view of interconnected assets across the plant. This has the potential to allow plant managers to move away from reactive and preventive maintenance to a predictive one, where essential machines don’t have to be brought down unless there is a definite abnormality.

At the foundational level this involves inserting sensors at key locations on the equipment for doing things such as vibration analysis of mechanical components like air compressors, belt drives or conveyors, fans and blowers, kiln rollers, motor bearings & vertical and horizontal mills to predict anomalies. We then collect the data coming out from these sensors, analyze and compute the triaxial vibrations, temperature, and noise of the equipment in real-time resulting in final equipment health score. A low score is sent to the plant manager, supervisor, or engineer with a diagnostic assessment of the probable cause for the irregularity and a recommendation for fixing the issue. If not severe yet significant, we then further monitor the fault with relevant parameters like temperature, vibration etc., to assure that the problem does not intensify and get worse. Emails, text messages or dashboards allow plant personnel to see the most relevant equipment data for the plant from single equipment access for a plant operator to multiple machines across the plant access for a plant head and a multi-plant machine score for the regional manufacturing head. 

Other common patterns that may indicate an issue are sudden changes in readings, spikes or dips in readings, and sustained changes over time. If there is a sudden change in readings from a sensor, this could be an indication that something is wrong. If there is a sustained change in readings over time, this could also indicate an issue. For example, if the temperature of a motor gradually increases over time, this could be an indication that the motor is overheating and may eventually fail. By analyzing data for these sorts of patterns, our AIs can often provide an early warning of potential problems so that maintenance can be performed before the issue becomes serious. If one of the sensors gives us readings that are consistently higher than normal, it is likely indicating an issue. Our AI can detect this pattern and raise an alert so that maintenance can be performed before the issue becomes serious. 

Besides sensors we also collect data in other ways:

-Operational data: This data includes information such as run times, feed rates, speeds, and pressures. 

-Maintenance data: This data includes information about past maintenance activities such as repairs, replacements, and inspections. 

We use #MAD to analyze this data to look for patterns that may indicate an impending failure. Machine learning allows our systems to learn from the data and improve over time. We use Machine learning algorithms to detect patterns in data that may indicate an impending failure. We then use a proactive approach that uses data to predict when an issue is likely to occur. 

Cloud providers too have got into predictive maintenance in a big way. Amazon has a service called Amazon Predictive Maintenance (AMP), Google has a predictive maintenance solution called Google Cloud Prediction API. Microsoft also has an Azure Machine Learning solution for predictive maintenance. All of these are cloud-based platforms that use machine learning to detect patterns in data. 

In the end what we are aiming for is to maximize the overall equipment effectiveness of our cement plants to improve equipment availability and performance, thus saving substantial costs for last minute or unwanted repairs and maintenance.

Supply Chain Optimization

Holcim has also used #MAD to optimize its supply chain. By analyzing data on customer demand, production capacity, and transportation logistics, the company can make informed decisions about where to source materials, how much to produce, and how to transport goods. This allows Holcim to minimize transportation costs, reduce lead times, and improve customer satisfaction by ensuring that products are delivered on time. As an example of scale, we work with over 20,000 transport suppliers and 90,000+ drivers that travel approximately 1.7 billion kilometers for us each year moving our material (source: Holcim Integrated Annual Report 2021). Holcim also uses data to forecast demand for its products, allowing the company to adjust its production capacity as needed to meet changing market conditions.

In our experience navigating our own AI journey within supply chain, we have identified four key principles for success:

1. Adopt an end-to-end perspective

2. Start small and scale fast

3. Enable people and machines to work together

4. Continuously learn and improve

By following these principles, companies can unleash the full potential of AI in their supply chains and transform them into self-regulating systems that continuously optimize performance. 

End-to-End Perspective

The first principle is to take an end-to-end view of the supply chain—from raw materials to finished products—and apply AI capabilities across the entire value chain. Most companies take a siloed approach, using AI to improve individual processes such as procurement, manufacturing, logistics, and customer service. A systemic view of the supply chain is necessary to identify and realize the full potential of AI. For example, by applying predictive analytics to demand planning, companies can generate more accurate forecasts of customer demand. This improved forecast can then be used to optimize production planning, which leads to reduced inventory levels and improved on-time delivery. In turn, lower inventory levels free up working capital, which can be reinvested in other areas of the business.

The benefits of an end-to-end perspective are not just theoretical. At Holcim we have achieved significant improvements in cost, service levels, and sustainability by adopting this approach. I was recently speaking to the CDO at a leading global consumer goods company that too was able to reduce manufacturing costs by 10 percent and inventory levels by 15 percent by applying AI across its end-to-end supply chain.

Start Small and Scale Fast

The second principle is to start small and scale fast. Many companies make the mistake of trying to transform their entire supply chain overnight. This “big bang” approach is rarely successful and often leads to project delays, cost overruns, and frustrated employees. A better approach is to start small—for example, by piloting AI in a single process or location—and then scaling up quickly to achieve broader benefits. This phased approach allows companies to learn and adapt as they go, which is essential for success in a rapidly evolving field like AI.

We followed this approach at Holcim and have successfully used AI to improve our material planning process. We started with a small pilot at one of our plants and then quickly scaled up the project to other plants. As a result of the initiative, we were able to reduce inventory levels and inventory turnover by a significant percentage.

Enable People and Machines to Work Together

The third principle is to enable people and machines to work together. Many companies view AI as a replacement for human workers, but this is not the best way to maximize its potential. In most cases, people and machines working together are more effective than either one working alone. We see this firsthand almost every day across our global markets in both the developed as well as developing nations. In our case by using AI to generate better insights and recommendations, our employees have been able to make better decisions about where to allocate resources. And by automating routine tasks, our staff can focus on higher-value activities such as developing new products or services.

One company that has also benefited from this approach is one of the largest US retailers, which uses AI to help its stores replenish inventory faster and more accurately. The system generates recommendations for store employees, who then make the final decision on what to order. As a result of the initiative, the retailer has been able to reduce out-of-stocks by a significant percent and increase sales by double digits as well.

Continuous Learning

The fourth and final principle is continuous learning. AI is a rapidly evolving field, and it is important for companies to keep up with the latest advances. One way to do this is to invest in research and development (R&D). This can be done internally or through partnerships with academia or other companies. We at Holcim work closely with the Massachusetts Institute of Technology as well as other institutions and have a significant global R&D arm in France. 

Another way to stay ahead of the curve is to adopt a “learning by doing” approach. This means using #MAD applications in the real world and then constantly iterating and improving based on feedback. For example, we have been testing machine learning algorithms to improve our product recommendations for our customers. We then then test these recommendations in the real world and fine-tune the algorithms based on how our customers respond.

We at Holcim have followed these four principles through a process of learning. It has helped fuel our AI efforts in managing our global supply chains. By following the above recommendations, we believe other companies too can make the most of AI in their supply chains and realize their full potential. 

Energy Efficiency

Holcim has used #MAD to improve its energy efficiency. The cement manufacturing process is energy-intensive, and the company has implemented numerous measures to reduce its energy consumption. By analyzing data on energy usage, Holcim can identify areas where it can reduce energy consumption, such as optimizing kiln operations or improving insulation. This has helped the company to reduce its carbon footprint and save money on energy costs. Additionally, by using data to monitor energy usage, Holcim can identify areas where it can further reduce consumption, such as by implementing more efficient lighting or HVAC systems.

Quality Control

Holcim uses data to monitor the quality of its products. By installing sensors and cameras throughout its facilities, the company collects data on various production parameters, such as temperature, humidity, and product size. This data is analyzed in real-time to ensure that products meet the highest quality standards. Additionally, by tracking production processes, Holcim can quickly identify and address any issues that may arise, reducing waste and improving efficiency. By using data to monitor quality control, Holcim ensures that its products meet customer expectations and comply with regulatory requirements.

Customer Analytics

Holcim has also used data to gain insights into customer behavior and preferences. By analyzing data on customer orders, complaints, and feedback, the company can identify patterns that indicate areas where it can improve. This has allowed Holcim to tailor its products and services to better meet customer needs, improving customer satisfaction and driving revenue growth. By analyzing data on customer behavior, Holcim can also identify emerging market trends and adjust its product offerings accordingly.

Data Visualization

Finally, Holcim has used data visualization tools to help managers make better decisions. By presenting data in a visual format, such as graphs or charts, managers can quickly identify patterns and make informed decisions based on data. This has helped Holcim to improve its decision-making processes and respond more quickly to changes in the market. By using data visualization tools, Holcim can also communicate data insights to stakeholders more effectively, ensuring that everyone in the organization has a clear understanding of the data.

In conclusion, Holcim has used #MAD - Machine Learning, AI and Data in a number of ways to improve its operations, reduce costs, and increase customer satisfaction. By leveraging data to predict equipment failures, optimize its supply chain, improve energy efficiency, monitor product quality, analyze customer behavior, and facilitate decision-making processes, Holcim has gained a competitive edge in the industry. Holcim's success with data is a testament to the power of data in driving business outcomes, and its data-driven approach has positioned us for continued success in the years to come.

If you’d like to learn more about what we are doing with #MAD please feel free to reach out. 

#ChooseDataForYourTeam