Xiaocheng AI
An AI assistant that truly understands the factory. It connects equipment and business data, understands processes and knowledge, and handles queries, scheduling, diagnostics and decisions in natural language, making complex production management as simple as a conversation.
Why is Line 3 running below target today?
Three main causes identified:
1. Changeover downtime of 45 minutes from 09:20 to 10:05, 18 minutes above standard;
2. Die M-2218 wear warning, raising the scrap rate by 0.9%;
3. Machine 5 is only 62% loaded and can take some work orders.
Recommendation: switch to die M-2231 and move WO-2085 to Line 5.
Four-layer architecture from data to decisions
Built layer by layer from the bottom up, so AI understands both real-time factory data and the implicit experience accumulated in the industry.
Application layer
Delivers actionable results for business roles, including a conversational assistant, smart reports, scheduling recommendations, equipment health dashboards and energy optimization, embedded seamlessly in MES and mobile.
Model layer
Combines large language models with domain-specific industrial algorithms such as scheduling optimization, anomaly detection and life prediction, balancing natural language understanding with industrial precision.
Knowledge base layer
Consolidates equipment manuals, process documents, SOPs, work order history and maintenance records into a company-specific industrial knowledge base, so AI answers are grounded in evidence.
Data layer
Unifies IoT equipment data, MES business data, ERP order data and energy data for real-time aggregation and governance of multi-source heterogeneous data.
Five core capabilities
Designed around real factory workflows; the sequence forms a complete loop from interaction to decision.
Unified voice and text entry
Bidirectional voice and text input lets shop-floor staff query, report and issue commands in one sentence. A unified entry integrates all system functions, so new employees can start the same day without memorizing menus.
- · Recognizes dialects and works in noisy environments
- · One sentence reaches the function, with no training required
- · Consistent across mobile, workstation terminals and PC
Intelligent support and fault diagnosis
Builds a knowledge base from equipment manuals, SOPs and maintenance history. On-site questions return illustrated handling steps immediately, and AI self-diagnosis accuracy for common faults exceeds 90%.
- · Full chain from symptom to cause to resolution
- · Maintenance knowledge accumulates, preserving veteran expertise
- · Alarm integration proactively pushes handling recommendations
Intelligent analytics and decision support
Automatically attributes fluctuations in output, yield and efficiency to specific lines, machines, shifts and work orders. Generate daily and weekly reports in one sentence and turn data into executable management actions.
- · Automatic anomaly attribution with drill-down to root cause
- · Natural-language business analysis reports
- · Proactive alerts on key metric fluctuations
Intelligent scheduling and dispatching
Generates schedules automatically from due dates, machine load, die availability and changeover time. Rush orders are rescheduled in seconds, raising on-time delivery to an average of 98%.
- · Multi-constraint optimization balancing delivery and equipment efficiency
- · Real-time rescheduling for rush orders keeps plans under control
- · Visual scheduling rationale lets planners fine-tune
Predictive maintenance
Builds equipment health models from vibration, current, temperature and pressure data, warning of potential failures 7 to 30 days in advance and cutting unplanned downtime by 45%.
- · Equipment health scoring and remaining useful life prediction
- · Early fault warnings shift maintenance from reactive to proactive
- · Spare parts demand forecasting for better inventory
Value for every role
One AI brain delivering dedicated value to every factory role.
See the business, focus on priorities
Get a plant-wide business report in one sentence, with automatic exception alerts, and see the real situation without waiting for layered reports.
Manage delivery, control risk
Track order progress and bottleneck operations in real time, with AI proactively flagging delivery risks and capacity gaps for early intervention.
Schedule faster and more accurately
Move beyond manual Excel scheduling. AI generates optimized plans in seconds, automatically reschedules rush orders and greatly reduces workload.
Repair earlier and more accurately
Equipment health warnings arrive early and fault diagnosis provides handling steps, turning firefighting into planned maintenance.
Easy to use, quick to learn
Voice reporting and visual work instructions remove the need to memorize complex menus. Ask AI when problems arise, so new staff get up to speed quickly.
Typical application scenarios
Covers high-frequency management scenarios from planning to execution and from equipment to energy.
Morning meeting data at a glance
When management asks about yesterday's output and exceptions by line, Xiaocheng AI immediately generates comparison data and an exception list, grounding the production meeting in facts.
Rush order response
When a rush order arrives, AI assesses the impact on the existing plan and proposes the fastest delivery option; one click confirms the reschedule.
Equipment exception handling
When equipment alarms, AI pushes likely causes and handling steps so maintenance staff can act directly, reducing downtime.
Energy anomaly tracing
When unit energy consumption rises abnormally at an operation, AI correlates output and equipment status, identifies the cause and recommends energy-saving actions.