Interpretation of the Connotations of Smart Factories and Digital Intelligence Utilization Rate
Smart Factories
A smart factory consists of “one entity and three dimensions”. The factory entity is composed of the equipment layer, control layer, workshop layer, factory layer, and enterprise layer. Currently, the overall architecture tends to be flat and cloud-based, with cloud-native technical architectures gradually maturing. The three dimensions are as follows:
Factory full-life cycle dimension: Covers stages such as factory design, construction, delivery, operation and maintenance, decommissioning, and recycling.
Product full-life cycle dimension: Includes links such as product R&D and design, process design, procurement, manufacturing, sales, service, disposal, and recycling.
Ecological dimension: Involves the supply chain, industrial chain, and value chain.
The entity serves as the foundation, and its function cannot be realized without each dimension. The construction level of the entity is determined by the factory’s full-life cycle management capabilities, while the operation level of the entity depends on the product’s full-life cycle capabilities and the factory’s ecological capabilities. Naturally, the effective operation of each dimension also relies on the support of the entity, and all dimensions are interrelated.
The construction level of a factory determines its basic capabilities in terms of operational efficiency, production flexibility, production safety, energy conservation and environmental protection, and digital intelligence. The operation level of the factory determines the product’s full-life cycle performance. The ecological level of the factory affects its adaptability and innovation capabilities in the supply chain, industrial chain, and value chain.
Digital Intelligence Utilization Rate
The underlying supporting elements of the “one entity and three dimensions” include humans, finances, materials, and data. As a new production factor, the realization of data value depends on the level of integrated data application. The level of integrated data application in a factory is determined by its digital intelligence utilization rate, which in turn decides the factory’s level of intelligence.
The digital intelligence utilization rate mentioned here is not a single indicator; it is jointly composed of data utilization rate and AI utilization rate:
Data utilization rate: Refers to the degree of effective use of data in a smart factory. Its foundation lies in the level of effective data collection, data governance, and data pipeline automation. Its core significance is to optimize the input-output ratio of resources, avoid “data hoarding”, and emphasize the practical utility of data in business scenarios.
AI utilization rate: Reflects the degree of value application of AI in a smart factory, focusing on how the system uses AI efficiently and systematically to create value.
Data-Driven Capabilities Determine Data Utilization Rate
Factories generate and collect massive amounts of data every day, covering all aspects of the “one entity and three dimensions” – such as market trends, customer needs, product sales, R&D, manufacturing, and operations. However, raw data is often disorganized and difficult to directly extract effective information from. The level of data governance and data pipeline automation capabilities is the foundation for improving data utilization rate, and data utilization rate is essentially determined by data-driven capabilities. Data-driven capabilities refer to the ability to optimize decision-making and support the achievement of business goals through data analysis.
In terms of the effectiveness of data-driven capabilities, it is mainly reflected in the degree of visualization and transparency of the factory:
Visualization helps grasp the factory’s operating status and identify problems.
Transparency reveals the root causes of problems and facilitates problem-solving.
The level of data visualization and transparency determines the extent to which data value is realized.
Visualization
Visualization is the basic form of data value realization. Its focus is to present key indicators, quantify business logic, identify business pain points, and thereby support process optimization and business decision-making. The in-depth goal of data visualization is to build an enterprise-level data analysis system and release and enhance data value based on business scenarios. For example, a machinery manufacturing enterprise improved the efficiency of identifying abnormal issues by 50% through the deployment of a visualization platform.
Transparency
Currently, transparency is mainly analyzed based on mechanism models, but in many scenarios, it can only rely on big data analysis, resulting in insufficient transparency. The gray-box model, formed by the integration of mechanism-based white-box models and big data-based black-box models, has unique advantages as it combines the internal interpretability of white-box models and the (relative) non-interpretability of black-box models. This model performs well in solving problems that are partially known and partially unknown: it retains the transparency of white-box models while inheriting the simplicity and efficiency of black-box models, providing a flexible, interpretable, and replicable solution. For example, a construction machinery enterprise increased the accuracy of equipment failure root cause analysis from 65% to 92% after introducing the gray-box model, effectively solving the dilemma of “knowing the result but not the reason”.
Data Intelligence Capabilities Determine AI Utilization Rate
With the development of AI technology, three key breakthroughs have been achieved through data governance and quality improvement, collaborative training of large and small models, and the combined application of AI with knowledge graphs and vector databases:
Extracting knowledge from raw data
Systematizing fragmented knowledge
Making implicit knowledge explicit
These breakthroughs further enable machine reasoning and the independent emergence of new knowledge. This technological evolution fundamentally solves the inefficiency in the generation, circulation, and application of knowledge, significantly improves the level of data application, promotes the shift from data-driven to data intelligence, and ultimately realizes the value release of data as a key production factor and artificial intelligence as a new productive force.
The level of data intelligence capabilities directly determines the actual application effectiveness of AI in factories. In factories, this capability is specifically reflected in two aspects:
The data foundation supporting AI development, including data scale, data richness, data quality, and knowledge density.
The application breadth and depth of multi-modal large models and intelligent agents that enable in-depth collaborative coexistence.
Large models provide cognitive and understanding capabilities for intelligent agents, while intelligent agents convert these capabilities into actual business value through a systematic architecture. The systematic architecture of intelligent agents designed based on the business characteristics of the factory will directly affect the application effectiveness of intelligent agents, thereby restricting the improvement of data intelligence capabilities. This is specifically reflected in two dimensions:
The Depth of Integration Between AI and Industrial Scenarios
For example, the integration of AI with industrial software such as ERP, PLM, APS, MES, and SCADA can be categorized into three levels:
“Plug-in calling”: AI tools operate independently of systems like MES, requiring manual work to close the loop.
“Platform embedding”: AI modules are integrated into systems such as ERP and APS to achieve intelligent enhancement and automatic data flow.
“AI-native reconstruction”: Reconstructing industrial software with AI as the core.
The Depth of Collaboration Between AI and Humans
This collaboration can be divided into three stages:
“Human-machine interaction”: For example, humans query AI analysis results through terminals.
“Human-machine collaboration”: For example, humans drive AI to adjust production management strategies, and AI assists humans in making management decisions.
“Human-machine symbiosis”: For example, humans focus on innovation and decision-making, while AI is responsible for implementation and execution.
Predictive Applications and Adaptive System Construction of Smart Factories
The effectiveness of data intelligence capabilities is ultimately reflected in two dimensions of the factory: predictive level and adaptive degree:
Prediction enables the anticipation of various situations in the factory, facilitating proactive responses. The predictive level is mainly reflected in the accuracy, timeliness, and application breadth of predictions.
Adaptation allows the manufacturing system to automatically adjust according to changes in demand and the environment. The adaptive degree is mainly reflected in the accuracy, real-time performance, and application breadth of adaptation.
Typical Predictive Applications of Smart Factories
In the future, predictions in smart factories will cover all aspects of the factory. Currently, the two most representative predictive applications are predictive maintenance of equipment and prediction of user needs.
Predictive Maintenance of Equipment
Real-time data (such as vibration, temperature, pressure, and current) is collected through sensors. In-depth mining of raw data is conducted to extract key features (in time domain, frequency domain, time-frequency domain, etc.) that reflect the equipment’s operating status. The remaining service life is predicted based on degradation trends, and early warning strategies are formulated in combination with business rules. AI has become an indispensable part of this process, and many equipment-intensive factories have built an “intelligent equipment brain”. For example, a wind power enterprise successfully reduced unplanned downtime by 60% by predicting the remaining service life of equipment through sensor data and AI models.
Prediction of User Needs
As the result of the comprehensive interaction of multiple factors (such as customer needs, market trends, and enterprise supply chain capabilities), AI prediction models can dynamically adjust decisions based on historical data and real-time demand to achieve intelligent recommendations, precise development, and timely production, thereby improving factory efficiency and response accuracy. For example, a fast-moving consumer goods (FMCG) enterprise increased the accuracy of demand prediction to 88% and improved inventory turnover by 35% through AI analysis of historical sales and market data.
Adaptive System Construction of Smart Factories
Future smart factories will be highly automated, flexible, human-machine symbiotic, and adaptive systems that can proactively adapt to changes in products and the environment. The data-driven and data intelligence capabilities based on the “one entity and three dimensions” ultimately determine the overall adaptive level of the factory. Factory adaptation mainly includes four aspects:
- Factory Entity Adaptation
It starts with the autonomous regulation of equipment, then progresses to the autonomous regulation of production lines, followed by workshops, and finally achieves full factory adaptation.
- Factory Full-Life Cycle Adaptation
Driven by data intelligence, it realizes continuous optimization and evolution covering all links of the factory’s full life cycle (design, construction, delivery, operation and maintenance, decommissioning, and recycling). It adapts to changes in the environment, market, and products, supports the smart factory in continuously adjusting itself under changing external conditions, and reflects the resilience and evolutionary capabilities throughout the full life cycle. It can quickly adapt to changes in product manufacturing needs and realize continuous factory innovation.
- Product Full-Life Cycle Adaptation
Supported by data intelligence, it realizes an intelligent optimization closed loop from sales, design, supply chain, production, logistics, delivery to service. It can adapt to changes in market demand and realize continuous product innovation.
- Ecological Adaptation
It connects external data sources such as the supply chain, industrial chain, value chain, and government, and establishes a data ecosystem across factories, industries, and fields. Combined with internal factory data, it forms a “global perspective”, identifies hidden correlations and new insights, promotes cross-boundary innovation, and builds a more three-dimensional manufacturing decision-making system.
Conclusion: Digital Intelligence Utilization Rate Determines the Competitiveness of Future Smart Factories
To summarize, data utilization rate is determined by data-driven capabilities, and AI utilization rate is determined by data intelligence capabilities; data drives decision-making, and intelligence reconstructs processes. From the perspective of factory development trends, digital intelligence utilization rate – i.e., the depth of a factory’s data-driven approach and the advancement speed of its data intelligence – has become the core criterion for distinguishing its competitiveness. In the future, only by continuously improving the digital intelligence utilization rate can enterprises build irreplaceable advantages in dimensions such as efficiency, flexibility, and innovation, and become leaders in the transformation of the manufacturing industry.
