AI4S Insight | Beyond "Isolated" Automation: Winning the Second Half of Catalyst R&D Through Workflow Collaboration
Your laboratory might appear to be "fully automated."
Materials are added automatically by a dosing platform, reactors operate unattended, and on-line chromatography enables real-time detection. All screens are active, and every piece of equipment is busily "working." Yet, when you begin developing a new catalyst, that familiar sense of helplessness returns. Progress remains hindered by slowness and disorganization. Pellet pressing must still be done manually, and evaluation tests require dozens of hours of continuous monitoring. Once results are available, you might find they contain abnormal data. Even worse, no one can clarify the exact process conditions under which that problematic sample was originally prepared.
While standalone instruments become increasingly expensive, the system workflow grows ever more fragmented.
Schematic Diagram of the High-Throughput Catalyst Synthesis Workstation Process
The problem isn't that individual instruments are too slow, but rather that there are disconnects between process steps. To turn a formulation into an evaluable catalyst sample, you must go through dosing, synthesis, drying and calcination, shaping, loading, and evaluation. A failure at any stage makes all upstream speed gains useless. This issue, which deserves top priority, cannot be solved by simply buying more equipment.
The Ou Shisheng high-throughput catalyst synthesis workstation is designed to bridge these gaps. It's not just another faster standalone instrument; instead, it connects the entire workflow seamlessly to create a truly collaborative "workstation."
I. The Hidden Bottleneck in Catalyst R&D: Automated“ Isolated Islands”
Over the past decade, individual unit operations in catalyst laboratories have indeed become automated. Dosing precision has improved, parallel reactors can run dozens of batches in one campaign, and evaluation workflows have been integrated with inline detection. However, assembling these “automation islands” reveals deeper discontinuities. Each instrument handles only a segment of the workflow, and the full chain has never been truly interconnected.
Data is fragmented between stages. Dosing records are logged in notebooks; pH adjustment during synthesis relies on manual empirical tuning; calcination requires on-site human monitoring; shaping depends heavily on manual dexterity; and sample loading is matched to specimens by memory. Data from each segment uses disparate formats and remains isolated. By the time samples reach the evaluation stage, the complete chain of process parameters can no longer be traced. After a set of experiments, what remains is not a data asset, but scattered fragments.
Post-processing constitutes an unautomated nonestaff stage. Critical steps that determine the final morphology and performance of catalysts, including pellet pressing, granulation, sieving and reactor loading, have long been omitted from automation solutions. Uneven scraped material residue causes tablet weight variation; inconsistent force in mortar granulation lowers shaping yield; manual sieving is prone to clogging and yields broad particle size distribution; bed packing density during loading depends entirely on operator experience, with missed or misloaded samples occurring from time to time. Every step introduces new errors, and all these operations remain manual with no automated backup.
Lack of unified standards renders results incomparable. Individual variability exists everywhere, from sample dispensing technique and pellet pressing pressure, to packing compaction and the timing of switching evaluation conditions. For the same formulation, researchers may obtain results differing by several percentage points in crystalline phase purity, tablet mechanical strength and bed homogeneity. Dispersed data and abundant variables make result comparison and reproducibility difficult across teams and institutions.
Throughput is the slowest step. The synthesis module can run dozens of batches in parallel, yet post-processing remains trapped in manual serial operations of “one pellet at a time, one tube tapped at a time.” Only one pellet can be pressed per cycle, and each reactor tube requires separate compaction during loading. The advantages of parallel synthesis for dozens of batches upstream are gradually eroded in post-processing. For the entire workflow, it still takes several days for a single sample from dosing to performance data acquisition, and the initial parallelization benefit is largely lost in downstream operations.
Safety risks and scale-up challenges compound each other. High-temperature calcination, strongly corrosive media, and toxic gas environments increase labor intensity and create safety hazards for researchers due to frequent exposure. Fragmented, unstructured manual records further deprive the scale-up process from lab to pilot scale of reliable process archives. Each scale-up attempt becomes a new round of trial and error.
Ultimately, the bottleneck is not that “a single step lacks automation,” but that “the full workflow is not automated end-to-end.” Each instrument handles only a portion of the process, forcing human operators to act as “connectors” between stages. This is the core problem that simply adding more equipment cannot resolve, yet it deserves top priority.
II. Ou Shisheng High-Throughput Catalyst Synthesis Workstation: Unifying the Workflow into a Continuous Pipeline
High-Throughput Catalyst Synthesis Workstation from Ou Shisheng
To address the long-standing challenge in catalyst R&D – "individual automated steps that cannot be connected together" – Ou Shisheng integrates the entire workflow into a single system. Through unified scheduling, it forms an automatic pipeline covering dosing, synthesis, drying and calcination, shaping, loading, and evaluation. Its core strengths can be categorized as follows:
Full-chain data connectivity. The unified visual scheduling software controls all workflow units for material management, batch scheduling, and process monitoring. It automatically correlates and traces process parameters across the entire chain, from synthesis to evaluation, building a closed-loop “process–performance” dataset. Full-process data is automatically acquired and logged, generating traceability reports. Cross-unit data is centrally managed with end-to-end traceability.
Catalyst Synthesis Platform (Including dosing unit, pH measurement unit, lid capping/decapping unit and reaction unit)
Full-process standardization and reproducibility. Solid dosing accuracy is ±0.1 mg, with static elimination and low-material warnings. Liquid dosing accuracy is ≤ ±1% FS, featuring automatic cleaning upon material switching and supporting 24-hour continuous dosing. Closed-loop inline pH control offers a precision of 0.01. Thirty-six parallel independent temperature-control channels (±1°C) and independent stirring are included. The reaction unit employs an intelligent temperature control strategy combining PID and fuzzy control, achieving temperature precision of ±1°C. A 100 Hz high-precision sensor array is deployed synchronously to capture multi-dimensional parameters of each channel in real time and dynamically correct deviations.
Automatic Pellet Pressing and Granulation Unit
Integrated post-processing. Four sequential operations—material scraping, pellet pressing, granulation, and sieving—are integrated. Pellet pressing pressure is adjustable from 0.5–5 MPa, achieving particle sizes of 40–60 mesh with a shaping yield greater than 75%. Automatic vibrating sieving and fractionated collection prevent clogging. Dual-station ultrasonic cleaning, along with rinsing and blow-drying, avoids cross-contamination; cleaning operates in parallel with production for uninterrupted continuous operation. Loading uses negative-pressure material suction. Column tubes are individually coded to match samples, enabling full-process traceability.
Matching pipeline schedule and unattended operation. Six-channel liquid preparation and 36-channel parallel reactions enable the synthesis of approximately 100 formulations within 24 hours. The oven features 9 chambers with temperature control to ±2°C, and the muffle furnace has 3 chambers with temperature control to ±1°C at high temperatures, including automatic door opening and closing. An Automated Guided Vehicle (AGV) transports materials between units, and programmed temperature control links synthesis, drying, and calcination. At the evaluation stage, on-line GC/MS performs automatic sampling, multi-dimensional data is calculated automatically, and column tubes are replaced automatically for continuous evaluation of multiple batches without interruption.
Safety control and end-to-end traceability. Programmed temperature control for drying, combined with intelligent air velocity management, shortens the cycle time by 30% and reduces energy consumption by 25%. The calcination module utilizes SiC heating and multi-point temperature measurement, achieving a temperature fluctuation of ≤±1°C at 1200°C. Remote monitoring and automatic alarms enhance high-temperature safety, while automatic doors and AGVs minimize human exposure to hazardous environments. Full-process data is automatically collected and archived, retaining complete structured records for scale-up research and technology transfer.
The synergy of these five strengths ultimately delivers four distinctive core capabilities of the Ou Shisheng fully automated high-throughput catalyst synthesis workstation: multi-chamber automatic drying and calcination, integrated pelletizing-granulating-cleaning, integrated catalyst loading-evaluation, and unified visual scheduling software. The first three fully connect the post-synthesis workflow after the reaction, while the fourth integrates individual units into one unified system-level platform. In terms of operational performance, the original serial workflow, which once took several days from dosing to obtaining performance data, is now compressed into an automated production line capable of 24-hour unattended continuous operation. Researchers can shift their core work from "operating instruments" to "analyzing data."
III. Generating Integrated "Process-Performance" Data for AI4S
The core principle of AI for Science is to enable machine learning models to learn rules from the relationship between "synthesis parameters and product performance," thereby predicting optimal formulations and process routes. However, the effectiveness of this approach hinges on the quality of the training data. The data set must be sufficiently large-scale and standardized, and contain complete process context, rather than isolated and fragmented experimental records.
Software Interface
The unified visual scheduling software connects all process units onto a single platform. It automatically collects and centrally records parameters, process information, and results, outputting structured datasets with well-defined fields that can be directly used for modeling. The software automatically correlates and traces process parameters across synthesis, calcination, shaping, loading, and evaluation. Each performance data point is linked to its full preceding process chain as a "process pedigree," establishing the data foundation for AI to perform feature selection and identify optimal process routes. Meanwhile, multi-dimensional process curves, such as temperature, pH, and stirring, are logged at a high sampling rate of 100 Hz (instead of only end-point values), enabling full root-cause traceback for any abnormal results. Combined with high-throughput capacity, including 6-channel liquid preparation, 36-channel parallel reactions, and the synthesis of approximately 100 formulations within 24 hours, data accumulation speed is greatly improved compared to conventional approaches, providing the prerequisite for training high-precision AI prediction models.
As structured data continuously accumulates and scales up, the software system evolves from "recording and management" to "decision support." Built upon the "process–performance" data set, pre-built models like Random Forest, XGBoost, neural networks, and symbolic regression can be directly invoked to rapidly construct surrogate models and predict the performance of new formulations. These models can also proactively recommend the next round of experimental plans to reduce blind trial-and-error. In this framework, hardware ensures high-throughput standardized execution; software builds a traceable data foundation; and AI extracts underlying patterns from the data, feeding back into the experimental closed loop. Researchers can then shift their core work from "operating instruments and organizing data" to "defining problems, analyzing models, and making scientific judgments."
IV. End-to-End Integration from Dosing to Evaluation: The Paradigm Shift of the Ou Shisheng Catalyst Synthesis Workstation
Accelerating catalyst R&D doesn't depend on adding another faster instrument, but on tightly integrating the entire workflow. When synthesis, shaping, loading, and evaluation become disconnected, what is lost is not just efficiency, but the entire asset of process data.
The Ou Shisheng automated high-throughput catalyst synthesis workstation integrates dosing, reaction, drying and calcination, shaping, loading, and evaluation into one traceable, intelligent production line. It transforms R&D from "experience-driven" to "data-driven," and from "assembled discrete instruments" to "full workflow collaboration."
As AI for Science rapidly advances, breaking through the complete data chain of "preparation—characterization—evaluation" and converting manual, experience-dependent craftsmanship into standardized, reproducible, and transferable engineering capabilities will be a critical breakthrough for the paradigm transformation of materials innovation. Centered on this objective, Ou Shisheng fulfills its mission and vision of "empowering scientists’ dreams through an intelligent instrument ecosystem." The company strives to incorporate every operational step of catalyst preparation and evaluation into a reliable, traceable, and collaborative intelligent instrument system. This represents not merely a generational upgrade of laboratory equipment, but a systematic reconstruction of the logic and methodology for catalyst research and development.
