Measurements for multi-scale optimization

WO2026217275A3PCT designated stage Publication Date: 2026-11-12SIC SYST
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/US2026/020133
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-03-21
Filing Date
2026-03-20
Publication Date
2026-11-12

AI Technical Summary

Technical Problem

Chemical and biomanufacturing processes face challenges in scaling up due to the difficulty in maintaining optimal reaction conditions, oxygen transfer limitations, mass transfer inefficiencies, and the need for cost-effective purification strategies, leading to inefficiencies in large-scale production.

Method used

The integration of generative artificial intelligence (AI) systems, including quantum generative AI models, to generate synthetic data and soft sensors, combined with anomaly-detection systems and agent-based AI, enables real-time optimization and control of manufacturing processes, bridging the gap between small-scale and large-scale operations.

Benefits of technology

This approach facilitates efficient scale-up of chemical and biomanufacturing processes, enhances process monitoring and control, reduces the time and cost associated with process development, and minimizes risks related to batch failures, while improving product quality and yield.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure US2026020133_12112026_PF_FP_ABST
    Figure US2026020133_12112026_PF_FP_ABST
Patent Text Reader

Abstract

In one aspect, in general, a method for autonomous optimization of a manufacturing process comprises: receiving at least one measurement at a production volume for the manufacturing process; applying a generative artificial intelligence (AI) system to generate synthetic data based on the at least one measurement; integrating at least one process model based on the synthetic data to generate a soft sensor that predicts a target measurement in the manufacturing process; and implementing an anomaly-detection system wherein the anomaly-detection system comprises at least one generator and at least one discriminator wherein at least one of the at least one generator and at least one discriminator are trained on an operation dataset.
Need to check novelty before this filing date? Find Prior Art

Description

Atty. Doc. No. SICS-100-C-WOMEASUREMENTS FOR MULTI-SCALE OPTIMIZATIONCROSS-REFERENCE TO RELATED APPLICATION(S)

[0001] This application claims priority to and the benefit of U.S. Application Serial No.19 / 087,319, entitled “MEASUREMENTS FOR MULTI-SCALE OPTIMIZATION,” filed March 21, 2025, the entire disclosure of which is hereby incorporated by reference.TECHNICAL FIELD

[0002] The disclosure generally relates to manufacturing, and in particular, methods and systems that use generative Al and / or soft sensors to optimize manufacturing processes.BACKGROUND

[0003] Chemical manufacturing processes, including biomanufacturing, often use various reaction systems, such as chemical reactors or cell-based biological systems (e.g., microorganisms and / or cell lines) operating in controlled environments to produce commercially valuable products useful in a wide range of industries, including pharmaceutical, agricultural, food, materials, energy, and specialty chemicals industries. Often, both biological and non-biological systems used in such processes are engineered or optimized to produce the desired product. Biomanufacturing is particularly critical for the large-scale production of commodity biologies, such as enzymes, amino acids, ethanol, organic acids, biofuels, biopolymers, and industrial proteins, which are used in applications ranging from food processing to bio-based materials and renewable energy. Chemical and biomanufacturing is used to produce a wide range of different types of products, including relatively simple organic molecules, like amino acids and lipids, more complex macromolecules, such as nucleic acids (DNAor RNA) and proteins, and synthetic chemicals, polymers, catalysts, and other complex compounds used in industrial applications. In some processes, the ultimate product may be a functional material or a structured biological entity, for example, when manufacturing therapeutic cells (e.g., as used in CAR-T cell therapy or embryonic cell therapy) or when producing engineered catalysts used in chemical synthesis.

[0004] One of the challenges associated with scaling up chemical and biomanufacturing processes to commercial-scale is the fact that maintaining optimal reaction or culture conditions may call for precise control over parameters such as temperature, pH, pressure, mixing, and substrate concentrations. For commodity biologies, large-scale fermentation and downstream processing introduce additional challenges, such as oxygen transfer limitations, mass transfer inefficiencies, and the need for cost-effective purification strategies.Optimization at large scale is often unfeasible, both because of the expense associated with large-scale reactors or bioreactors and the difficulty in collecting critical data from industrialscale production systems. Because of this, optimization is typically performed using smaller- scale laboratory or pilot plant setups. However, conditions that appear optimal at small scale often do not translate effectively to larger-scale production systems, resulting in the failure of many scale-up attempts.SUMMARY

[0005] In one aspect, in general, a method for autonomous optimization of a manufacturing process comprises: receiving at least one measurement at a production volume for the manufacturing process ; applying a generative artificial intelligence (Al) system to generate synthetic data based on the at least one measurement ; integrating at least one process model based on the synthetic data to generate a soft sensor that predicts a target measurement in the manufacturing process ; and implementing an anomaly-detection system wherein the anomaly-detection system comprises at least one generator and at least one discriminator wherein at least one of the at least one generator and at least one discriminator are trained on an operation dataset.

[0006] Aspects can include one or more of the following features.

[0007] The manufacturing process is a biomanufacturing process.

[0008] The manufacturing process is a chemical manufacturing process.

[0009] The manufacturing process is a commodity biologies manufacturing process.

[0010] The generative Al system comprises a quantum generative Al model.

[0011] The method further comprises using the soft sensor to generate target data in the manufacturing process.

[0012] The target data is used to monitor the manufacturing process.

[0013] The target data is used for a control system of the manufacturing process.

[0014] The method further comprises generating a benchmark dataset of time-series process data comprising normal operating regimes and anomalous operating regimes, wherein the benchmark dataset is produced by a dynamic simulation model of an integrated upstream and downstream system, and wherein the operation dataset is derived from the normal operating regimes of the benchmark dataset.

[0015] The anomaly-detection system comprises at least one generative adversarial network comprising: a generator neural network configured to receive a latent vector sampled from a quantum probability distribution and map the latent vector to synthetic process data ; and a discriminator neural network trained to distinguish the synthetic process data from data associated with the normal operating regime.

[0016] The method further comprises: storing the trained discriminator neural network of the at least one generative adversarial network; applying input time-series data stream representing measurements from the manufacturing process into the trained discriminator neural network; computing an anomaly score for the trained discriminator neural network ; and labeling the input time-series data stream as anomalous when the discriminator neural network scores the input time-series data stream outside a learned distribution of normal operation.

[0017] In another aspect, in general, a method for autonomous optimization of a manufacturing process comprises: receiving at least one measurement at a production volume for the manufacturing process; applying a generative artificial intelligence (Al) system to generate synthetic data based on the at least one measurement ; integrating at least one process model based on the synthetic data to generate a soft sensor that predicts a target measurement in the manufacturing process; and implementing an agent-based Al system structured as an ensemble of semi-autonomous agents to perform real-time inference and control, wherein the agent-based Al system utilizes the soft sensor and the synthetic data to execute an agent-based decision process for optimizing at least one condition of the manufacturing process.

[0018] In another aspect, in general, a method for quantum optimization of a manufacturing process comprises: receiving at least one measurement at a production volume for themanufacturing process ; applying a quantum generative artificial intelligence (Al) system to generate data based on the at least one measurement ; integrating one or more process models based on the generated data ; predicting a dynamic of the production volume using a soft sensor ; and optimizing a condition in a large-scale manufacturing process via an agent-based Al system, wherein the agent-based Al system comprises an ensemble of semi-autonomous agents that utilize the dynamic to execute an agent-based decision process for real-time autonomous control

[0019] Aspects can have one or more of the following advantages.

[0020] Aspects can have one or more of the following advantages.

[0021] Some embodiments of the techniques described herein enable the efficient scale-up of chemical and biomanufacturing processes, particularly for the large-scale production of commodity biologies and other high-volume chemical products, and / or that facilitate the collection of critical data from large-scale reactors, bioreactors, or other production systems.

[0022] Some embodiments of the techniques described herein provide significant advantages in manufacturing process optimization by leveraging generative artificial intelligence (Al) to enhance process monitoring, control, and scalability. The disclosed methods address traditional challenges by utilizing Al-generated synthetic process measurements in conjunction with soft sensors, enabling real-time data-driven decision-making. This approach allows for the estimation of critical process parameters that are otherwise difficult to measure, thereby improving overall process transparency and control. By integrating generative Al with quantum optimization techniques, some of the techniques described herein facilitate iterative, cyclical optimization processes that dynamically refine manufacturing conditions based on real-time and synthetic data, leading to improved efficiency and yield in large-scale manufacturing environments.

[0023] Another potential advantage of the disclosed techniques is the ability to bridge the gap between small-scale experimental data and full-scale manufacturing operations. By applying generative Al to synthesize process data and predict optimal conditions for large- scale operations, the present disclosure enables a more seamless transition from development to production. This allows manufacturers to reduce the time and cost associated with process development, while also minimizing risks related to batch failures, inconsistent product quality, and suboptimal process conditions. Furthermore, the integration of advanced controlsystems informed by Al-driven models ensures robust process monitoring and adaptive control strategies, making it possible to proactively adjust operating parameters in response to process fluctuations. Some embodiments of the techniques described herein incorporate an ensemble-based scoring mechanism within an anomaly-detection system to identify systemic anomalies while minimizing false positive alerts.

[0024] Other features and advantages will become apparent from the following description, and from the figures and claims.BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The disclosure is best understood from the following detailed description when read in conjunction with the accompanying drawings. It is emphasized that, according to common practice, the various features of the drawings are not to-scale. On the contrary, the dimensions of the various features are arbitrarily expanded or reduced for clarity.

[0026] Fig. 1 is an exemplary process diagram, in accordance with one or more embodiments of the disclosure.

[0027] Fig. 2 is an exemplary illustration of an embodiment of the present disclosure, in accordance with one or more embodiments of this disclosure.

[0028] Fig. 3 is a process flow diagram, in accordance with one or more embodiments of the disclosure.

[0029] Fig. 4 is an upstream process diagram, in accordance with one or more embodiments of this disclosure.

[0030] Fig. 5 is an downstream process diagram, in accordance with one or more embodiments of this disclosure.

[0031] Fig. 6 is an open-loop system response, in accordance with one or more embodiments of this disclosure.

[0032] Fig. 7 is a closed-loop system process scheme, in accordance with one or more embodiments of this disclosure.

[0033] Fig. 8 is a graph illustrating response time, in accordance with one or more embodiments of this disclosure.

[0034] Fig. 9 is an exemplary Simulink sensor model.

[0035] Fig. 10 is an exemplary actuator model, in accordance with one or more embodimentsof this disclosure.

[0036] Fig. 11 illustrates a framework for conducting parameter analysis within a biomanufacturing process, in accordance with one or more embodiments of this disclosure.

[0037] Fig. 12A is an illustration of an exemplary embodiment of a QT-Tube combining a microtube and a macrotube, in accordance with one or more embodiments of this disclosure.

[0038] Fig. 12B is an illustration of quantum computing and Al in QT-Tube, in accordance with one or more embodiments of this disclosure.

[0039] Fig. 13 is a diagram of an exemplary quantum computer, in accordance with one or more embodiments of this disclosure.

[0040] Fig. 14A is a flowchart of an exemplary quantum computing process, in accordance with one or more embodiments of this disclosure.

[0041] Fig. 14B is a schematic of a quantum and classical computer process, in accordance with one or more embodiments of this disclosure.

[0042] Fig. 15 is a schematic of a quantum and classical computer, in accordance with one or more embodiments of this disclosure.DETAILED DESCRIPTION

[0043] Reference will now be made in detail to the exemplary embodiments of the present disclosure, examples of which are illustrated in the accompanying drawings. Wherever possible, the same reference numbers will be used throughout the drawings to refer to the same or like parts.

[0044] It is to be understood that although the invention has been described above in terms of particular embodiments, the foregoing embodiments are provided as illustrative only, and do not limit or define the scope of the invention. Various other embodiments, including but not limited to the following, are also within the scope of the claims. For example, elements and components described herein may be further divided into additional components or joined together to form fewer components for performing the same functions.

[0045] The systems, devices, and methods disclosed herein are described in detail by way of examples and with reference to the figures. The examples discussed herein are examples only and are provided to assist in the explanation of the apparatuses, devices, systems, and methods described herein. None of the features or components shown in the drawings ordiscussed below should be taken as mandatory for any specific implementation of any of these devices, systems, or methods unless specifically designated as mandatory.

[0046] Also, for any methods described, regardless of whether the method is described in conjunction with a flow diagram, it should be understood that unless otherwise specified or required by context, any explicit or implicit ordering of steps performed in the execution of a method does not imply that those steps must be performed in the order presented but instead may be performed in a different order or in parallel.

[0047] As used herein, the term “exemplary” is used in the sense of “example,” rather than “ideal.” Moreover, the terms “a” and “an” herein do not denote a limitation of quantity, but rather denote the presence of one or more of the referenced items.

[0048] In certain aspects, provided herein are methods and systems that use generative artificial intelligence Al to determine optimized large-scale fermentation conditions from data generated in small-scale culture systems. In some aspects, also provided herein are methods and systems for using generative Al to generate synthetic process measurements in large-scale systems. Such use of soft sensors (also known as virtual sensors) to generate other measurements not able to be easily physically measured allows for optimization of large scale manufacturing. The combination of synthetic data generation and quantum optimization can be used to make iterative, cyclical optimization processes.

[0049] The pharmaceutical industry's shift towards biological therapeutics has led to a transition from conventional batch production to continuous manufacturing. This change highlights the crucial need for effective process monitoring and control strategies to ensure consistent product quality and stability. Open-source benchmark simulation models have become essential tools for refining these processes, offering a platform for testing research hypotheses, serving as educational resources, and assessing plant performance.Biomanufacturing encompasses various upstream and downstream unit operations, including cell culture, harvesting, cell disruption, and product formulation. A comprehensive dynamic model covering both upstream and downstream components provides an integrated perspective of the production process. An advanced control system emphasizing realistic sensor and actuator integration is used to enhance simulation accuracy. The control system assesses the benchmark through open-loop and closed-loop simulations, offering an in-depth analysis of the model's dynamic response and functionality. Embodiments of the presentdisclosure represents a model-driven decision support tool that enables in silico evaluation of monitoring strategies, process design, process optimization, and control for biomanufacturing.

[0050] Batch processes find extensive application in biologies manufacturing, facilitating the synthesis of diverse bioproducts encompassing pharmaceuticals, enzymes, and biopharmaceuticals. Nevertheless, they are accompanied by a distinct array of complexities such as: ensuring uniformity in product quality across sequential batches, preserving aseptic conditions within facilities to forestall contamination, grappling with protracted cycle durations relative to their continuous counterparts that could impede production efficacy and protract time-to-market intervals, necessitating substantial allocations of resources, spanning water, energy, and labor, exhibiting diminished adaptability vis-a-vis continuous methodologies, thereby hindering swift adjustments in tandem with market demand fluctuations, and engendering elevated waste volumes compared to continuous processes represent pivotal challenges inherent to batch processes. Addressing these challenges routinely mandates a synthesis of comprehensive scientific comprehension, adept engineering prowess, cutting-edge process surveillance and regulation technologies, and unwavering commitments to iterative enhancement endeavors.

[0051] In some biopharmaceutical processes, the workflow is divided into two phases:Upstream and Downstream. The upstream phase involves cell cultivation and harvesting, while downstream processing encompasses a series of activities from cell separation and / or disruption to the formulation of the final purified product. This process is pivotal not only in biopharmaceuticals but also in chemical and commodity biologies. Downstream processing steps include centrifugation for biomass cell separation, filtration to obtain a concentrated biomolecule stream and purification treatments. The distinction between these phases is particularly significant in biopharmaceuticals, where the ultimate goal is the production of a purified and high-quality therapeutic. The adoption of batch processes of this nature often stems from the expansion of bench-scale production techniques. The final product must conform to stringent quality regulations established by regulatory authorities, and separating each stage facilitates effective quality monitoring. Nonetheless, driven by escalating global competition, the industrial landscape is progressively transitioning towards a standard of continuous manufacturing.

[0052] Bioprocess and chemical engineering requires information at various levels of granularity across scales like metabolic pathways, reaction kinetics, reactor specifications, and interaction of the micro to the macroscale and vice versa. While many of these phenomena can be modeled mechanistically, engineering towards the goal of maximizing process throughput and other constraints satisfaction requires tools from artificial intelligence (Al) such as search, optimization, sampling, and machine learning to model process correlations. The complexity is further enhanced due to unavailability of data from many critical microscale states such as metabolic fluxes, cell morphology, viability, etc. not disregarding the stochasticity in the dynamics of the individual cells and observability limitations of the process. Developing systematic algorithms validated by experiments forms the foundation to develop pipelines for accelerated scale-up of biomolecules.

[0053] There is a lack of transferability of measurement technologies and scale-up strategies in bioprocessing and chemical manufacturing. It is also computationally intensive to estimate the state of a cell in a reactor given just the operating conditions of the reactor. As such, there is a need to develop methods of estimating the state of a reactor from macroscopic measurements and process causations obtained from controlled experiments, and to use process information to trigger pathways to aid in manufacturing desirable compounds, thereby avoiding byproducts and inhibitors.

[0054] Embodiments of the present disclosure are presented to address the following needs in biomanufacturing:1. Synchronization of Upstream and Downstream Processes: embodiments of the present disclosure offer a novel, plant-wide model that integrates upstream and downstream processes based on industrial guidance, literature review, and thermodynamic principles. This integration enables users to assess the impact of critical upstream process parameters on the final product, enhancing understanding and optimization possibilities.2. Transparency, Availability, and Flexibility: embodiments of the present disclosure include an open-source model, freely accessible to the scientific community via GitHub. A baseline scenario is provided with initial values, constant variables, parameters, and assumptions, allowing for transparent evaluation of upstream, downstream, and unit operation effects. Additionally, embodiments offerflexibility for further research or adaptation to new host organisms.3. Mathematical Model for Process Design, Control, and Monitoring: embodiments of the present disclosure introduce a novel basis control strategy informed by literature, industrial perspectives, and mathematical indices like the Niederlinski index. Through comprehensive examples, embodiments demonstrate the efficacy of these control strategies in managing critical process parameter variations. The inclusion of sensor and control models, accounting for time lag and disturbances, enhances realism and applicability.4. Case Study framework: embodiments of the present disclosure include a tailored case study framework to facilitate research on continuous biomanufacturing. This framework guides users through experimentation and analysis, offering insights into process optimization.

[0055] Advanced monitoring from imaging, spectroscopy, and capacitance probes, have made it possible to have at-line / on-line monitors. While great instrumentational advancements, sporadic measurements and transferability of results post-calibration are a hindrance. However, some online instrumentation and independent experiments aided by first principles can be leveraged to generate explainable monitoring solutions. One aim is to create a transferable real-time data-agnostic monitoring system, with process knowledge gained from micro and macroscale experiments by integrating a diverse range of offline sensors that can help in generating fast scale-up solutions.Biomanufacturing

[0056] For the purpose of illustration, some examples will be described in the context of biomanufacturing as an example of a manufacturing process that benefits from the disclosed systems and methods. However, it should be understood that the techniques described herein are not limited to biomanufacturing. Rather, the techniques may be broadly applicable to any type of chemical manufacturing, biomanufacturing, or commodity biologies production, as well as other related industrial processes that require similar operational improvements.

[0057] In certain aspects, provided herein are methods and systems related to the use of generative Al in the monitoring and / or optimization of manufacturing processes. In certain embodiments, the biomanufacturing processes disclosed herein utilize biological systems (e g., microorganisms, such as bacteria and yeast, animal cells, including human cells andcell lines, and / or plant cells) to produce bioproducts used in a range of industries. In certain embodiments, the products of such biomanufacturing can be primary metabolites produced by microorganisms (e.g., ethanol, acetone, butanol, amino acids, organic acids such as lactate, acetate, citrate), secondary metabolites produced by microorganisms (microorganism- derived antibiotics, such as penicillin, tetracycline, streptomycin), macromolecules (e.g., nucleic acids (DNA, RNA), lipids, polypeptide / proteins, including protein-based drugs and enzyme-based biocatalysts), synthetic biology products (e.g., antimalarial drug artemisinin by engineered E. coli and yeast, isobutanol from engineered E. coli, and opioids by engineered yeast). Biomanufacturing disclosed herein can also produce cells, including stem cells (e.g., embryonic stem (ES) cells, induced pluripotent stem cells (iPSCs), immune cells (T cells, B cells, NK cells, CAR-T cells, CAR-NK cells) for regenerative medicines and immunotherapies. Additional examples of biomanufacturing processes and products can be found in Zhang et al., "Biomanufacturing: history and perspective." Journal of Industrial Microbiology and Biotechnology 44.4-5 (2017): 773-784, the contents of which is hereby incorporated by reference in its entirety.

[0058] In some biomanufacturing processes, the workflow can be divided into two phases:upstream and downstream. The upstream phase involves cell cultivation and harvesting, while downstream processing encompasses a series of activities from cell separation and / or disruption to the formulation of the final purified product. Downstream processing steps include centrifugation for biomass cell separation, filtration to obtain a concentrated biomolecule stream and purification treatments.

[0059] In some biomanufacturing processes disclosed herein, microorganisms, cells, and their genetically engineered counterparts have been used as “mini-factories” in biomanufacturing to aid with the diversity of metabolic pathways and their accompanying ability to transform a wide range of renewable raw materials into value-added compounds by means of fermentation. Successful examples include Escherichia coli, which produce insulin and carotene, Saccharomyces cerevisiae, which produce geraniol, Yarrowia lipolytica, which produce N-acetylneuraminic acid, Bacillus subtilis, which produce hyaluronic acid, and so on.

[0060] For most biomanufacturing processes, fermentation in a bioreactor is the core unit, where the raw materials are converted into a target product with the catalysis by enzyme(s) orbiological cells. Fermentation can be performed in batch, fed-batch or continuous mode. Since the 1950s, batch or fed-batch fermentation process has been the standard design and practice for biomanufacturing of fuels, chemicals, and high-value products due to its advantages in simple bioreactor structure and design, easy scale-up, and flexibility for different products. In the past few decades, continuous fermentation has been emerging as a new approach to significantly increase the productivities and reduce the capital investment and biomanufacturing cost, but contamination risks remain a major challenge of this approach. Continuous processes may also require continuous monitoring of parameters that necessitates an array of real-time sensors such as temperature sensors, level sensors and residual solvent levels that require periodic maintenance and calibration, contributing significantly to operational costs. Xie, Dongming. "Continuous biomanufacturing with microbes — upstream progresses and challenges." Current opinion in biotechnology 78 (2022): 102793, the contents of which is hereby incorporated by reference in its entirety.

[0061] Fermentation systems have complex cellular metabolic behavior and complex hierarchical structures. These systems consist of cells (e.g., microorganisms) and a fermentation environment, and they are influenced by upstream and downstream operating conditions. Fermentation may be aerobic, microaerobic, or anaerobic. Fermentation processes used for biomanufacturing generally involve genetically modified (recombinant) cells and microorganisms. Various genetic modification technologies developed to target certain formation products can be found in U.S. Patent U.S. 11,098,307, the contents of which is hereby incorporated by reference in its entirety. Fermentation media used with recombinant microorganisms and cells may contain suitable carbon sources or substrates for both growth and production stages. Suitable substrates may include, but are not limited to glucose, sucrose, xylose, mannose, arabinose, oils, carbon dioxide, carbon monoxide, methane, methanol, formaldehyde and glycerol. It is contemplated that, in certain embodiments, all of the above mentioned carbon substrates and mixtures thereof are suitable as a carbon source(s). In addition to an appropriate carbon source, such as selected from one of the herein-disclosed types, fermentation media may contain suitable minerals, salts, cofactors, buffers and other components, known to those skilled in the art, suitable for the growth of the cultures and promotion of the enzymatic pathway necessary for chemical product fermentation. Typically, cells are grown at a temperature in the range of about 25°Cto about 40°C in an appropriate medium, as well as up to 70°C for thermophilic microorganisms. Suitable pH ranges for the fermentation are between pH 2.0 to pH 10.0, where pH 6.0 to pH 8.0 is a typical pH range for the initial condition. However, the actual culture conditions for a particular fermentation are not meant to be limited by these pH ranges. Suitable growth media are well characterized and known in the art and can be found in U.S. Patent U.S. 11,098,307, the contents of which is hereby incorporated by reference in its entirety. In certain embodiments, the methods and systems provided here can facilitate the optimization of culture conditions in large scale fermentation processes.

[0062] Exemplary information regarding the scale-up fermentation in manufacturing is described in Thiry et al., "Optimizing scale-up fermentation processes." TRENDS in Biotechnology 20.3 (2002): 103-105, the contents of which is hereby incorporated by reference in its entirety. During industrialization, the expansion of the bioreactor volume and changes in shape lead to changes in the fermentation environment, which often leads to the failure of fermentation strategies that are developed during the laboratory stage. The factors affected by scale include, but are not limited to, the number of generations, the mutation probability, medium sterilization, the quality of temperature and pH regulations, agitation, aeration, and pressure. Often, the scaling-up of a process is facilitated by first scaling-down to the pilot scale of the conditions of culture that are contemplating being used at the final scale of production. Then, when the scale is increased, the broth will become more and more heterogeneous. The fermentation development strategy requires knowledge of the behavior of the cells and its bioproducts, which can be facilitated using certain embodiments of the methods and systems provided herein. In certain embodiments, the fermentation process is developed taking into account the feedback from the downstream process (DSP) development, the final scale of production, and the good manufacturing practice requirements. In some instances, even subtle changes to culture conditions can have significant effects on biomanufacturing productivity. In certain embodiments, the methods and systems provided herein facilitate prediction of the behavior of large scale cultures based on pilot scale culture behavior, which can in certain instances be pivotal to the success of the scaling up the process.

[0063] Scaling-up fermentation processes from the laboratory to industrial production level requires significant level of optimization and control, as discussed in Du, Yuan-Hang, et al.,"Optimization and scale-up of fermentation processes driven by models." Bioengineering 9.9 (2022): 473. The physiology of cells and microorganisms and the flow field in the bioreactor are two important issues that are highly interrelated throughout the fermentation process, and they affect the final fermentation yield. In addition, in industrial bioreactors, the spatiotemporal gradients of substrate, temperature, and pH can affect the metabolism of the resulting strains. Strong agitation in industrial bioreactors may also cause shear damage to cells and reduce their viability. These factors can lead to a loss of performance in scale-up operations, including a reduction in titer, yield, or productivity, as well as potential pitfalls for process monitoring and product quality control.

[0064] Most biomanufacturing processes include the cultivation of cells and microorganisms, which implies a process consisting of complex chemical, physical and biological phenomena. A dependable and consistent analytical system is necessary to control the process conditions in all parts of the biomanufacturing process (upstream, downstream and product formulation). For example, the upstream processing part, once it includes cell growth, is a complicated multi-phase system with a considerable variability due to the inherent nature of the cell cultures. Hence, sensors are required to measure parameters of biomanufacturing processes.

[0065] In the particular case of fermentation processes, measuring products such as cells, proteins and by-products is crucial to guarantee that the production process is developing as expected. The opportunity to assess how the fermentation is progressing makes it possible to not only optimize the process by promptly regulating specific parameters, but also to spot when the process is not under optimal operation. As such, additional measurements may be used to measure variables such as dissolved CO2 (in the off-gas), optical density, and cell density. In the case of cell density, it is usually measured using optical in-situ probes, or chemometrics coupled with intact cell mass spectrometry (ICMS), impedance spectroscopy, near infrared spectroscopy (NIR) or Fourier transform-infrared spectroscopy (FTIR).

[0066] Monitoring methods and their associated sensors and analyzers can be generally categorized according to their position regarding the process unit as in-line, at-line or offline. Gargalo, Carina L., et al., "Towards smart biomanufacturing: a perspective on recent developments in industrial measurement and monitoring technologies for bio-based production processes." Journal of Industrial Microbiology & Biotechnology: Official Journalof the Society for Industrial Microbiology and Biotechnology 47.11 (2020): 947-964, the contents of which is hereby incorporated by reference in its entirety.SOFT SENSORS

[0067] As provided herein, soft sensors are advanced process monitoring systems, which use algorithms to assess measurements in an on-line manner to generate information about an otherwise unmeasured process state. In some embodiments, on-line measurements may include measurements provided by probes, sensors, devices, or instruments in or in the vicinity of a bioreactor or any other bioprocess unit operation where the signals are capable of just-in-time monitoring. In certain embodiments, soft sensors enable the monitoring of critical conditions that indeed affect the bioprocesses but cannot be measured directly in realtime. For example, in some embodiments, the soft sensors provided herein use (relatively) easily accessible on-line data to determine other cultivation variables that are otherwise either difficult to measure or can only be measured at low frequency. Alternatively, soft sensors can also be used within process monitoring, e.g., with the purpose of performing fault detection and diagnosis. Soft sensors can be generally categorized into two different classes, namely model-driven and data-driven.

[0068] Embodiments of the present disclosure include the integration of quantum hardware (and more generally, quantum generative artificial intelligence (Al)) for creating synthetic data in fermentation processes. These techniques represent advancement over classical generative Al models. For example, the inherent ability of quantum computing to handle complex, multidimensional data sets allows for the generation of highly detailed and diverse synthetic datasets that more accurately mimic the nonlinear, dynamic nature of chemical and biological fermentation processes. For example, the fed-batch culture begins with a batch culture. After the exponential growth phase (i.e., a period in which the number of new bacteria appearing per unit time is proportional to the present population), different substrates and media components might be added continuously to the fermenter at specific rates to provide nutrition and maintain a suitable environment for the cells’ growth. The above processes are repeated until the end of the final batch phase, resulting in the nonlinear, dynamic nature of the fermentation processes. This quantum advantage enables the exploration of larger parameter spaces and the modeling of fermentation dynamics with a level of precision and depth not feasible with classical computing methods. The capabilitiesof classical generative AT are inherently limited by the computational complexities and the exponential scaling of problem sizes. Quantum generative Al, in contrast, can leverage quantum superposition and entanglement to efficiently solve complex optimization problems and generate synthetic data that may lead to breakthroughs in predicting fermentation outcomes, optimizing process parameters, and enhancing the overall understanding of bioprocesses.Using Generative Al to Generate Synthetic Bioprocess Measurements

[0069] Synthetic data, generated through computer simulations based on mathematical models of chemical and biological processes such as fermentation, allows researchers and engineers to explore the impacts of various parameters on different outcomes without the need for extensive, costly, and time-consuming experimental trials. By simulating different operational conditions synthetic data helps identify optimal process conditions, predict performance outcomes, and guide decision-making and decision-support with a significantly reduced experimental burden.

[0070] Synthetic data also plays a crucial role in training and validating soft sensors and machine learning models in situations where real-world data may be scarce, expensive to obtain, or biased due to limited operational conditions. The generation of synthetic data that accurately reflects a wide range of biological process dynamics enables the development of robust models capable of predicting performance under diverse conditions. This is particularly useful in enhancing process monitoring and control strategies, leading to improved product quality, yield, and process efficiency. For example, in predictive maintenance, models trained on synthetic data can foresee equipment failures or process deviations, allowing for preemptive actions to be taken before actual issues arise. Ultimately, the strategic use of synthetic data accelerates the adoption of advanced analytics and automation technologies in manufacturing, fostering innovation and efficiency improvements while mitigating risks associated with process scale-up and commercialization.

[0071] Embodiments of the present disclosure include the integration of quantum hardware (and more generally, quantum generative artificial intelligence (Al)) for creating synthetic data in chemical and fermentation processes. These techniques represent advancement over classical generative Al models. For example, the inherent ability of quantum computing to handle complex, multidimensional data sets allows for the generation of highly detailed anddiverse synthetic datasets that more accurately mimic the nonlinear, dynamic nature of biological fermentation processes. This quantum advantage enables the exploration of larger parameter spaces and the modeling of fermentation dynamics with a level of precision and depth not feasible with classical computing methods. While classical generative Al has been utilized in simulating chemical and fermentation processes, its capabilities are inherently limited by the computational complexities and the exponential scaling of problem sizes. Quantum generative Al, in contrast, can leverage quantum superposition and entanglement to efficiently solve complex optimization problems and generate synthetic data that may lead to breakthroughs in predicting fermentation outcomes, optimizing process parameters, and enhancing the overall understanding of such processes.

[0072] One embodiment for generating such synthetic data utilizes a Generative Adversarial Networks (GAN). Traditional GANs consist of two main components: a generator whose role is to create data that is indistinguishable from real data, and a discriminator whose role is to differentiate between the real data and the data generated by the generator. This adversarial setup encourages the generator to produce increasingly realistic data over time. The training process involves back-and-forth optimization where the generator tries to fool the discriminator, and the discriminator tries to become better at identifying fake data, leading to improvements in the quality of the generated data.

[0073] Tensor networks, such as those discussed in Lopez -Piquieres et. al and herein incorporated by reference, can be used to train the GAN, as well as for optimization purposes. In some embodiments of the present disclosure, arbitrary integer-valued equality constraints of the form Ax = b, are encoded directly into U(l) symmetric tensor networks (TNs) and leverage their applicability as quantum-inspired generative models to assist in the search of solutions to combinatorial optimization problems within a generator-enhanced optimization framework. This allows for the exploitation of the generalization capabilities of TN generative models while constraining them so that they only output valid samples. A constrained TN generative model efficiently captures the constraints by reducing number of parameters and computational costs. At tasks with constraints given by arbitrary equalities, symmetric matrix product states outperform their standard unconstrained counterparts at finding novel and better solutions to combinatorial optimization problems.

[0074] Another embodiment for generating such synthetic data utilizes quantum associativeadversarial networks (AAN), such as the quantum-classical AAN described by Anschuetz et.al or Arici, both herein incorporated by reference in their entirety. An associative adversarial network (AAN) is a conceptual extension of GAN, designed to capture and leverage associative relationships within data for enhanced generative modeling. AANs use additional neural network layers that connect the discriminator to the generator; in quantum AANs, this additional layer is typically executed on quantum hardware.Using Soft Sensors to Generate Measurements Not Easily Physically Measured

[0075] Traditionally, physical sensors in a fermentation tank have been necessary for understanding and optimizing the process and efficiencies of large scale manufacturing. Knowing the conditions in a reactor or tank allows a manufacturer to optimize the conditions for reaction speed and output. Physical sensors typically measure quantities such as optical density, pH, dissolved oxygen, temperature, conductivity, and carbon dioxide evolution. However, it is likely that a user wants to know more about the conditions in areactor in order to further optimize the process. For example, the following factors, if available in real-time, may be invaluable in augmenting and optimizing that process:1. Direct measurement of specific metabolites;2. Real-time enzyme activity;3. Intracellular parameters;4. Continuous culture viability and morphology;5. Real-time genetic expression levels;6. Detailed biomass content;7. Concentration of trace elements;8. Heterogeneity and distribution of biomass inside the reactor; and 9. Diffusion coefficients and nutrient solubility in fermentation media.

[0076] However, many of these measurements are unviable using physical sensors because these sensors are constrained by hardware limitations and costly setups, not to mention changing sensors is potentially disruptive to the manufacturing process itself. In many cases, even when the data can be collected with physical sensors, time delays due to the analyses render such data unusable for the purposes of real-time control and optimization.

[0077] Soft sensors are a technology designed to estimate unmeasurable or difficult-to- measure process variables in real-time, leveraging available process data and mathematicalmodels. Soft sensors synthesize data from multiple physical sensors and incorporate that data into algorithms or models to infer values of parameters that are otherwise challenging, expensive, or invasive to measure directly. Soft sensors are extensively used in industries such as chemical manufacturing, pharmaceuticals, environmental monitoring, and biotechnology, offering a cost-effective solution for enhancing process control, optimization, and quality assurance by providing real-time data and insights that are otherwise inaccessible through conventional sensing technologies.

[0078] Embodiments of the present disclosure include utilization of different sensors ranging from spectroscopic, imaging, off-line and at-line experiments to ensure the viability of the resultant product.

[0079] In some embodiments, sensors are utilized to collect various data, including, but not limited to, physical variables (e.g., temperature, light intensity, pressure, shear stress distribution), chemical quantities (e.g., pH, dissolved oxygen, dissolved CO2, substrate concentrations, spatiotemporal gradients of substrate), as well as biological parameters (e.g., cell density, cell morphologies, concentrations of metabolites, and concentrations of cell secretions such as total protein output), and changes thereof over time.

[0080] In some embodiments, depending on the particular metabolic pathways of a microorganism, sensors may be used to detect and / or monitor cellular production of fermentation product(s) other than the desired fermentation product, selected from the group consisting of acetate, acetoin, acetone, acrylic, malate, fatty acid ethyl esters, isoprenoids, glycerol, ethylene glycol, ethylene, propylene, butylene, isobutylene, ethyl acetate, vinyl acetate, other acetates, 1,4-butanediol, 2,3-butanediol, butanol, isobutanol, sec-butanol, butyrate, isobutyrate, 2-OH-isobutryate, 3-OH-butyrate, ethanol, isopropanol, D-lactate, L- lactate, pyruvate, itaconate, levulinate, glucarate, glutarate, caprolactam, adipic acid, propanol, isopropanol, fused alcohols, and 1,2-propanediol, 1,3-propanediol, formate, fumaric acid, propionic acid, succinic acid, valeric acid, maleic acid and polyhydroxybutyrate.

[0081] In some embodiments, sensors can also be used to control the transition between growth and production in multi-stage fermentations by detecting and / or monitoring the concentration depletion of one or more limiting nutrients that are consumed during growth, including, but not limited to, phosphate, nitrogen, sulfur and magnesium.

[0082] In some embodiments, at least one in-line sensor is used to produce data continuously (no sampling), and it is in direct contact with the process medium (invasive) or separated from the medium by a glass window (bypass, non-invasive, also called on-line sensor). By providing continuous information, these sensors are enablers of continuous process control. In some embodiments, data collected by in-line sensors include, but not limited to, temperature, pH, level of off-gas, air, oxygen, CO2, reaction pressure, viscosity, and stirrer speed. To collect these data, various in-line sensors can be used, such as near-infrared spectroscopy (NIR), dielectric spectroscopy (DS), Fourier-transform infrared spectroscopy (FTIR), fluorescence (FS), Raman spectroscopy.

[0083] In some embodiments, at least one at-line sensors is used to analyze samples in close proximity to the reactor. Even though the samples are collected at regular time-intervals (manually or automatically), time delays due to the analysis (depending on the equipment) render such data suitable for monitoring purposes, but not for control. Examples for at-line sensors include proton transfer reaction mass spectrometry (PTR-MS), matrix-assisted laser desorption ionization (MALDI-TOF-MS), time-of-flight mass spectrometry, nuclear magnetic resonance (NMR), flow-injection analysis (FIA), flow cytometry, and so on. In some embodiments, the data collected by at-line sensors allow monitoring concentration of various components of the fermentation media such as biomass, glucose, lactate, amino acids, enzymes, antibodies, organic acids, vitamins, recombinant proteins, volatile organic compounds.In some embodiments, samples for off-line measurements are collected manually or automatically and then transferred to the laboratory to be analyzed. The analyses cause long time delays, so such measurements can contribute to the control of the dynamic process behavior data suitable for monitoring purposes, but not for control. In some embodiments, the analysis process can involve using instrumentation as high-performance liquid chromatography (HPLC), enzyme-linked immunosorbent assay (ELISA), gas chromatography (GC), mass spectrometry (MS), electrophoresis, microscopy.

[0084] Fig. l is a process diagram 10 of a technique for predicting a target measurement in a manufacturing process. The manufacturing process may comprise a biomanufacturing process, a chemical manufacturing process, or a commodity biologies manufacturing process. In step 12, the process may include receiving at least one measurement at small-scaleproduction volumes for the manufacturing process. Measurements may comprise time-series data of inline sensors and concentration prediction coming from different operating conditions generated via high-fidelity models. The small-scale production volumes may comprise multiple volumes between 1 microliter and 500 liters. In step 14, the process may include applying a generative artificial intelligence (Al) system to generate synthetic data based on the at least one measurement. The generative Al system may be either classical or quantum. In some implementations, a classical generative Al system may comprise a generative adversarial network (GAN) or associated adversarial network (AAN). In some implementations, a quantum generative Al system may comprise a system with actual quantum hardware, such as a Quantum-GAN or Quantum-AAN, or a quantum-inspired network such as a tensor network (TN) utilizing classical hardware. In step 16, the process may include integrating at least one process model based on the synthetic data to generate a soft sensor that predicts the target measurement in the manufacturing process. The process may further output a recreation of time-series data of both inline and synthetic variables like concentration variability in the reactor, temperature, pH, etc. In some implementations the soft sensor may be used to generate target data in the manufacturing process.

[0085] In some implementations, the manufacturing process of process diagram 10 may further utilize at least one reactor having at least one condition. In some implementations, the reactor may comprise a bioreactor or a chemical reactor. In some implementations, the at least one condition may comprise a fermentation condition, an antiseptic condition, a reaction condition, or a culture condition. In some implementations, the condition may comprise parameters such as temperature, pH, pressure, mixing or substrate concentrations, dissolved oxygen, conductivity, or carbon dioxide evolution. In some implementations, in step 16, the soft sensor may predict the target measurement based on the measurement of the at least one condition.

[0086] In some implementations, the manufacturing process of process diagram 10 may further comprise receiving at least one scaled-up measurement from a scaled-up manufacturing process wherein the at least one measurement from the scaled-up manufacturing process includes measurements collected using the soft sensor. In some implementations, the scaled-up manufacturing process is at a scale of greater than 1000 liters. In some implementations, the at least one scaled-up measurement is used to optimize the atleast one condition. In some implementations the soft sensor may be used to generate target data in the manufacturing process. In some implementations, the target data may be used to generate a predicted dynamic of the scaled-up manufacturing process. In some implementations, the target data may be used to monitor the manufacturing process. In some implementations, the target data may be used for a control system of the manufacturing process. In some implementations, the control system may include utilizing an agent-based decision process.

[0087] The advent of advanced sensor systems and big data integration combined with generative Al and agent-based Al systems allows for much more granular and autonomous process control and real time optimization of complex systems. Advanced process modeling, agent-based and generative Al models have emerged as powerful tools to address the challenges of complex non-linear systems. Agent-based Al refers to Al systems that function autonomously within a defined framework to make real-time inferences and decisions. These systems are structured as ensembles of semi -autonomous agents, each responsible for specific inference and control tasks. The agent-based Al framework enables the integration of generative models, sensor data, and mechanistic process knowledge to optimize and control complex industrial systems. By leveraging agent-based architectures, such Al systems enhance process automation, enable real-time monitoring, and improve decision-making capabilities in industrial applications. This approach is particularly effective in managing non-linear and dynamic systems where direct measurement of critical variables is challenging, thus providing a robust and adaptive solution for process optimization and control.

[0088] Fig. 2 illustrates an exemplary overview of an embodiment of the model. The model consists of an upstream, downstream, and control structure. The upstream has been designed to maximize productivity, the downstream has been designed to isolate the product from the other components in the stream and the basic control structure has been designed to monitor and control the unit operations and the synchronization. In essence, the model can be considered a pioneering open-source, plant-wide benchmark simulation model for the biopharmaceutical industry. Its availability empowers researchers and practitioners to explore deeper into process optimization, control, and monitoring, thereby transforming continuous biomanufacturing practices and addressing longstanding challenges in the field.Exemplary Model Process DescriptionHost Organism Specifications

[0089] In an exemplary embodiment, Lovastatin, belonging to the statin class of drugs, is synthesized. Lovastatin is primarily synthesized by Penicillium Citrinum, Monascus Ruber, and Aspergillus Terreus. In the case of Aspergillus Terreus, lovastatin is produced as a secondary metabolite, exhibiting extracellular expression in the form of P-hydroxy acid, also known as mevinolinic acid (referred to as lovastatin). This production process is characterized by relatively low productivity, typically around 6 mglllh under optimal conditions during fed-batch operation. Users of the model can change the host organism to investigate the optimal condition for the case study simulation.Process Configuration: Upstream and Downstream

[0090] The model includes a comprehensive framework for both upstream production and downstream purification of the target biomolecule, which is produced in a fermentation process. As illustrated in Fig.3, the initial phase of the upstream process involves the introduction of the essential raw materials, namely Adenine and Lactose, into a mixing unit designated as M-101. The resultant media, after undergoing thorough homogenization, proceeds in a continuous manner towards a Continuous Stirred-Tank Reactor (CSTR) represented as R-101. Within the confines of this fermenter, the biomass, primarily consisting of Aspergillus Terreus, initiates the production of lovastatin. Following this, the fermentation broth, derived from the CSTR fermenter (R-101), undergoes subsequent treatment through a hydro-cyclone separator denoted as HC-101. This hydro-cyclone assumes a pivotal role in the separation process. Specifically, it yields two distinct output streams: a rich stream (R) characterized by a substantial biomass component. A portion of this rich stream is cyclically reintroduced into the CSTR, effectively contributing to the ongoing fermentation process. In contrast, the other stream, notably less abundant in terms of biomass (denoted as stream 5), is channeled downstream for further processing.

[0091] The downstream segment of the process encompasses a series of sequential separation procedures. The initial stages involve centrifugation steps, denoted as C-101 and C-102, which serve the purpose of effectively eliminating residual biomass and other dense impurities from the solution. Subsequently, the process advances to a nanofiltration stage designated as NF-101, engineered to achieve solution concentration prior to the ultimatepurification phase. During the nanofiltration step, the concentration of lovastatin is achieved through a mechanism based on steric exclusion. This process allows the retention of lovastatin molecules while permitting the passage of smaller components. The subsequent phase involves a chromatography segment, comprising a set of six reversed-phase columns identified as LC-101 through LC-106. This chromatography section is meticulously devised to yield a solution enriched with purified lovastatin molecules. Fig. 3 serves as an illustrative representation of the process flow diagram governing the production of the Active Pharmaceutical Ingredient (API) lovastatin, and includes an upstream module aimed for increasing Lovastatin production (V represent valves, R represents reactor, P represents pump and HC represents hydro cyclone), downstream process aimed for Lovastatin purification and related control structure for dynamic process control (C represents centrifuge, T represents tank and NF represents nanofiltration)Plant-wide Model

[0092] The adopted model finds its foundation in investigations conducted on the liquid suspension growth of Aspergillus Terreus ATCC 20542 strain. Within this context, Lactose (LA) serves as the primary carbon source, facilitating both the growth of biomass (X) and the synthesis of Lovastatin (LO). Concurrently, yeast extract (N) plays a pivotal role as the primary nitrogen source. The subsequent rate equations can be formulated to characterize the consumption kinetics of lactose (HA) and yeast extract (rx), in addition to the kinetics governing the generation of biomass (rx) and lovastatin (FLO):

[0093] The optimal feed concentrations are determined from experimental runs. A criticalaspect pertains to the sterility of the feed, whereby the absence of biomass is a fundamental prerequisite throughout the entirety of the fresh feed. The process of mixing is indispensably executed to attain a uniform feed stream for the CSTR. This stage involves the integration of the recycle stream and the fresh feed, with the presumption of perfect and instantaneous mixing for the purposes of this model. The entirety of the production process is envisaged to occur within the confines of the CSTR, incorporating the assumption of a consistent density profile over time and the absence of liquid accumulation. Temperature emerges as another pivotal parameter in this context. To sustain the desired temperature, heat exchange is typically facilitated through reactor heating or cooling jackets. Based on the reference elucidated for the ongoing reaction, the magnitude of heat generation within this fermentation process is considered negligible. Consequently, a constant temperature profile is assumed for the simulation.

[0094] Dissolved oxygen (C02) is an example of a nutrient used by microorganism for growth, maintenance and metabolic production which can be describe as follows:Oxygen Transfer Rate (OTR) = KLA(C^.2— C&,s) (5) Oxygen Uptake Rate (OUR) =

[0095] In the computation of oxygen concentration within the fermentation reactor, pivotal elements encompass the determination of oxygen consumption corresponding to lactose utilization yield, the assessment of volumetric mass transfer coefficient, and the establishment of oxygen saturation concentration. The evaluation of oxygen consumption per lactose consumption yield adheres to reaction stoichiometry. The volumetric mass transfer coefficient, a determinant of gas-liquid mass transfer, is contingent upon variables including agitation intensity, aeration rate, substrate properties, temperature, pH, and environmental context. Notably, a constant KLA value is assumed, according to reported values. The relevant parameters governing dissolved oxygen modelling are presented in Table 1, to calculate oxygen concentration which impacts fermentation efficacy and microbial metabolism.Table 1: Dissolved oxygen modelling parameter

[0096] The oxygen transfer rate is determined by the product of the volumetric mass transfer coefficient and the difference between the dissolved oxygen concentration) and the oxygen saturation concentration (CQ2). The oxygen uptake rate, on the other hand, quantifies the oxygen consumption during the growth of biomass.

[0097] The integration of the hydro-cyclone into the recirculation loop introduces notable advantages, encompassing heightened productivity and streamlined downstream processing. However, it comes with the trade-off of heightened operational costs due to the necessary additional pumping. In establishing an optimally designed process flow rate, the model proposed by (Nemeth & Verdes, 2011) assumes prominence. This model is based on utilization of the split ratio denoted as a, where a = Qc / Qs Another significant parameter enters the fray: the separation efficiency, represented as 0, mathematically defined as ? = Cx,c / CX,4-

[0098] Post the hydro-cyclone stage, the presence of a purge stream serves the vital purpose of regulating the extent of recycling. This component is indispensable to prevent the accumulation of by-products and non-viable biomass within the recycling loop. Furthermore, critical design parameters come into play within the continuous fermentation system featuring biomass recycle. These parameters encompass the volumetric recirculation factor (as per Eq. 8), which establishes the flow rate of fresh substrate feed (QR), contingent upon the flow rate of recycled biomass (Qi). The cell recirculation ratio (as per Eq. 9) establishes a correlation between the biomass concentration in the recycle stream (CXR) and the concentration exiting the bioreactor CX 3). Lastly, the purge ratio (as per Eq. 10) governs the flow rate of the purge stream (QP).F_ 2(iC"Oc

[0099] The impact of the recirculation factor is investigated through iterative analyses involving variations in the reactor inlet flowrate across a range of trial recirculation factor (RF) values. This investigation is conducted while maintaining the remaining upstream parameters constant. The ultimate criterion for selecting the recirculation factor is contingent upon identifying the value that corresponds to the most pronounced production peak.Following this analytical pursuit, the optimal recirculation factor is found to be 0.8.

[0100] Fig. 4 is an upstream section diagram including CSTR reactor for fermentation (R- 101), pump (P-101), valves (V-101-103) and hydro cyclone (HC-101) for separation of biomass which is partially recycled and purged to increase the reaction conversion and avoid of the accumulation in the process.

[0101] The primary objective of the downstream system design, Fig. 5, revolves around the isolation of the Active Pharmaceutical Ingredient (API) molecule from the fermentation broth. While the upstream section design emphasizes maximizing productivity, the downstream focus shifts towards achieving the utmost sustainable capacity of the separation processes to gain both separation and the minimum design capacity.

[0102] The choice of employing centrifugation (C-101 & C-102) as the initial step is common within the pharmaceutical industry. This approach facilitates the effective elimination of any solid particulates that could potentially hinder subsequent separation processes. A pivotal factor in the selection of a centrifuge revolves around the concept of radial force, typically quantified in terms of gravitational force. In the context of the employed model, this radial force (F) is interconnected with the distance from the center of rotation (r) and the rotational velocity (N). This relationship is mathematically expressed through the equation:

[0103] To estimate the quantity of separated cells using a straightforward approach, aseparation efficiency factor (X) has been implemented. The selection of the separation efficiency factor is informed by an extensive literature review, with reported values ranging from X = 90% to 98%. Notably, centrifugation alone is deemed insufficient if the biomass concentration surpasses the specified threshold of 30 mg / L. To address this, embodiments of the design strategy involve employing two centrifuges in series (C-101 and C-102), as indicated in the process diagram. This design decision aligns with the implemented steadystate mass balance and maintains the assumption of constant density as utilized in the upstream model. Additionally, the analysis assumes that biomass concentration is the sole parameter influenced by this particular step.

[0104] Buffer tanks play a pivotal role in continuous manufacturing as an efficient method for flow rate control and avoiding snow-ball effects in recycle stream loops. Although in continuous operation, buffer tank capacities can often be substantially minimized, and in some cases even eliminated, their inclusion remains relevant for stability and redundancy considerations. The tank's capacity is intricately linked to its level of filling. Notably, despite process fluctuations being confined to the range of 1-5 L / h, even modest deviations can disrupt the optimal functioning of separation units. An additional purpose of tank T-101 is to facilitate filling during nanofiltration membrane maintenance downtime, a critical step in maintaining consistent separation performance. Given Lovastatin's vulnerability to pH- dependent hydrolytic degradation, minimizing residence time in an unregulated environment is imperative. Drawing from analogous instances in the literature, an implemented tank capacity of 500 liters is rationalized.

[0105] Elevating the concentration of API Lovastatin within the clarified broth yields both cost reductions and enhanced capacity for preparative chromatography. Nanofiltration (NF- 101) commonly stands as the preferred choice for fermenter broth concentration, particularly within the pharmaceutical sector. Filtration is driven by the pressure gradient across the membrane, with certain components being retained due to the membrane's physical attributes and geometry, while others can pass through it. With a molecular weight of 404.5 g / mol, Lovastatin necessitates a membrane boasting a molecular weight cut-off of 300 to 500 g / mol. For molecules of this magnitude, pressure gradients generally span between 5 - 50 bar, accompanied by an average pore size of 0.5 - 2 nm. In the context of mass transfer, the implemented model considers two crucial parameters. The first is related to the solvent'spermeate through the membrane (PE), while the second centers on the rejection factor of the i solute (Rej ,i), denoting the ratio between the component's concentration in the permeate stream and its concentration on the feed side ( f).

[0106] In this context, a pressure-dependent rejection factor has been employed for Lovastatin. This approach enables the utilization of the pump-generated pressure gradient to regulate nanofiltration performance to a certain extent. Concerning solvent permanence through the membrane, the permeate volumetric flux (Jw), estimated via the adapted Hagen- Poiseuille equation tailored to membrane parameters, has been adopted. The permanent volumetric flux and applied values for distinct parameters are outlined as follows:Table 2: Applied value for different parameter

[0107] Tank (T-102) serves as an accumulator, collecting incoming volume before facilitating its transfer to the chromatography section equipped with multiple parallel columns to ensure a semi -continuous flow. This tank maintains a constant fdling process and discharges a specific liquid volume at regular intervals. Given the extended retention time of the chromatography operation, the tank assumes a pivotal role. Furthermore, the tank acts as a means to alleviate minor fluctuations in feed concentration prior to injection. The ultimate purification step involves chromatographic columns (HPLC-101). The employed method is reversed-phase High-Performance Liquid Chromatography (RP-HPLC), which can be broken down into the following steps:1. An equilibration solution is introduced through the column to establish the nominal physical-chemical properties of the stationary phase.2. The sample, under high pressure, is injected through the column packed with stationary phase components.3. Due to distinct interactions with the stationary phase, the retention of various components within the sample solution is uneven, enabling differentiation between each molecule. Those with the lowest interaction proceed first.4. An elution solvent courses through the column to recover molecules bonded with the stationary phase. This solvent is typically stored in a designated tank (T-103), with a predefined volume injected for each elution.5. Subsequently, the eluted substance is subjected to spectrophotometric analysis to determine the composition and manage a fractionating valve.

[0108] Fig. 5 is a downstream section diagram including two step centrifuging based on the input concentration of biomass and chromatography column respectively aimed for the separation of biomass, Lovastatin, and Lactose. Embodiments of the present disclosure employ the reversed-phase High-Performance Liquid Chromatography (RP-HPLC) method. In reversephase chromatography, the mobile phase is comparatively polar, while the stationary phase is non-polar. As a result of more robust chemical interactions, non-polar compounds exhibit extended retention times. The species mass balance applied is as follows:

[0109] Where macc,i is mass accumulation term, mConv,i is convective mass flowrate, mdtsp,i is dispersed mass flowrate and, mmt,i is mass transfer into particle. It is assumed that the total mass transfer into the particle surface is equal to the accumulation of the ith component. macc,ads,t is overall accumulation of the ith component in the stationary phase.

[0110] The formulation and quantity of columns were designed under specific limitations. It is presumed that the flow rate of the injected solvent remains constant at Qs = 20 Uh , and the duration of solvent injection is recuperated through the subsequent method:

[0111] Here ttnj represents the injection time in hours, and Vinj stands for the volume of theinjected solvent. The capacity of each column volume is established at 15 L, and the residence time for every column operation is set at 12 h. Consequently, to prevent liquid accumulation it is necessary to dimension a chromatography system capable of handling the accumulated volume in tank T-102 within a 12-hour timeframe. The quantity of sample for each injection is 30.8 liters, distributed evenly among a grouping of six parallel chromatographic columns. The collective injection time for the devised system spans 1.54 hours. Utilizing this information, the flow velocity of the mobile phase can be estimated:Qs (17)u=T - A HPLC

[0112] Where AHPLC is section area of HPLC column and u is mobile phase flow velocity. The final column operative parameters and column geometrical dimension are as follows:Table 3: HPLC applied parameters

[0113] In Table 4, sets of independent equations of the process sections are provided. It may be convenient to summarize the variables for each stream, as a reference for the following control design. Table 5 represents the individual variables of the continuous sections.Table 5: Sets of independent equation for exemplary processesTable 5: Independent variables of continuous sectionDynamic Performance

[0114] Following representative process description and model establishment, various simulation scenarios were considered. Firstly, an evaluation of the open-loop system has been undertaken to comprehend the simulator's response prior to any closed-loop control consideration. Two distinct scenarios were examined to assess the open-loop system's dynamic response in the face of input fluctuations. Subsequently, a simulation analysis was conducted for both the upstream and downstream components, with the aim of developing a default control system. This was achieved through a comprehensive degree of freedom analysis, supplemented by insights drawn from relevant literature, real-world industrial cases, relative gain arrays, and the Niederlinski Index. To enhance the validity of the dynamic simulation control model, a control handle was incorporated into the closed-loop system. This addition accounts for both time lag and noise in the measurements, contributing to a more accurate portrayal of the system's behavior. Finally, two distinct scenarios were introduced to illustrate the performance of the KTB1 control system in the face of varying input conditions.Open Loop Assessment

[0115] A series of stepwise changes were carried out to verify the response of the isolated models. The trial specifications and their corresponding outcomes are shown in Fig. 6. All simulations were executed to attain a stable state, with no control systems in operation. In the initial trial, Impact of 50% increasing the organic nitrogen source load in the reactor feed has been investigated, which resulted in a proportional rise in organic nitrogen concentration within buffer tank T-101. This change also led to an 8.5% increase in API concentration,highlighting the significant role of the organic nitrogen source in lovastatin synthesis. It is important to note that API productivity is very low, which results in a highly diluted fermentation broth. By adjusting the input parameters within a continuous benchmark, it is possible to achieve optimal and suboptimal compositions for the fresh feed.

[0116] The second trial focuses on the 25% augmentation of the carbon source in the reactor feed. This adjustment results in a proportional increase in lactose concentration within T-101. Notably, the lovastatin concentration exhibits a 16% increase compared to the base case. It can be reasonably inferred that the heightened lovastatin content is constrained by the presence of an excessive gap between substrate quantities, leading to an inhibition in lactose consumption for lovastatin synthesis. In conclusion, this underscores the substantial impact of the inlet composition on lovastatin productivity.Simulation Analysis and Set-Up Default Control System

[0117] In the upstream process, equilibrium equations and the kinetic model were created using logical blocks. To enhance the reliability of the simulation, the integration of a delay block is considered, causing a temporal lag in the qualities of the recycling stream denoted as R. The delay is established as:

[0118] The technique works by storing input values and relative time steps in an internal buffer. If necessary due to lack of data, the block automatically extrapolates to the closest set of available values.

[0119] This upstream model is implemented as an optimal steady state. Each tuned input is tuned for maximum achievable efficiency and stable performance. So in this case, the goal of the control structure is to keep the operating conditions very close to the reference case. This decision is also made due to the nature of the process, given that many changes in the pharmaceutical process are not permitted.

[0120] The controller of this system has been designed based on three concepts, (i) Based on a degree of freedom (DOF) analysis to analyze which manipulable variables are available to control the process are identified. The DOF is defined as the number of controlled variables that can be regulated by the control system; (ii) relative gain array (RGA) is then analyzed to find the control loop pairing with the least interactions; (iii) Niederlinski Index (NI) is used to find the possible control loop pairing.

[0121] All upstream process units such as mixer, bioreactor, and hydro-cyclone rely on a general volume balance and four species balance equations. When these equations are supplemented with the purge volumetric balance, the total number of equations reaches sixteen. By rearranging and correcting the equations and applying all simplifications, the three hydro-cyclone equation become null identities. Therefore, the number of equations is reduced to 13. On the other hand the number of independent variable according to table 6 with consideration of reactor volume should be 16. Therefore, DOF is 3 indicating the number of manipulated variables. In this respect, process stream no.l, stream no.2 and purge stream volumetric flow (QI, Q4 and Qp) are selected as manipulated variables and respectively recycle factor (RF), reactor volume (V) and stream 4 lovastatin concentration (CI0V,4) are selected as controlled variable.

[0122] Fig. 7 is a closed loop process schematic of an exemplary embodiment in Simulink, including different units and related control structure.

[0123] The models of each downstream operative unit in can be implemented in MATLAB / Simulink (Mathworks, USA), with the balances presented previously. As seen in the upstream section, the control system's goal is to maintain operating conditions close to the previously restored optimal state. It is worth noting that in continuous manufacturing, the performance of the downstream section is highly dependent on the flow rate and composition of the stream from upstream. If the upstream section is not regulated correctly there is a limited set of actions that can be performed to reduce the production loss.

[0124] The control design procedure for the downstream section, due to the higher complexity of the unit operations, follows a different path. A trial-and-error strategy may be applied for the tuning of control system and then disturbance is introduced to the model to demonstrate the effectiveness of the control strategy. PI controller for C-102 is designed for the centrifuge separation by increasing its rotational speed. PI controller is implemented for T-102 to regulate the buffer volume with an extremely smooth action. This is done with the objective of preventing process fluctuations from impacting the nanofiltration NF-101 operation. For NF-101 a feed pump control provides the required pressure gradient. For obtaining stable lovastatin concentration in the retentive stream a PI controller has been designed to provide the suitable rotational speed of VSP-101 regarding the nanofiltration process. T-102 precedes the chromatography section and delimitates the continuous part ofthe process. Therefore, a discrete-time P control system is designed to discharge a constant volume at regular time intervals, based on the injection periods of the chromatography section. The tank volume is 100 L, while the total injection time equals 1.54 h. Therefore, the control system must be designed to bring the T-102 liquid volume back to 1001 before the next injection. Fig. 7 shows the complete simulation scheme in Simulink with the control structure.

[0125] For downstream and upstream processes, the parameter of each controller is selected based on analytical tuning rules (the SIMC rules applied here). The model, control structure and process units have been implemented into a computational simulation model. The controller parameter for the related controller to downstream and upstream has been mentioned in Table 6 respectively.Table 6: Upstream and Downstream controllers and tuned parameter valueSensor Model and Control Handles

[0126] One of the most important aspects of designing and modelling a control system is considering a realistic sensor and actuator model. In the dynamic simulation of embodiments of the present disclosure, the ideal and simplified sensor models are represented as a major parameter for comparability with field application, appropriate sensors and actuators make dynamic model more realistic.Sensors

[0127] Sensor response time and measuring noise are two major parameters in the realistic description of the sensor model. Response time has a definition based on step response as presented in Fig. 8. The normal response time consists of delay time and rise (or fall) time. The delay time (T10) is the time to reach 10% of the final value of a step response. The overall time to reach 90% of the final value of the step response is defined as response time (T90).

[0128] It is challenging to comprehensively consider noise effects but for more realistic modelling some of the noise effect has been considered in embodiments of the present disclosure. The noise signal is white zero-mean normally distributed with a standard deviation of 1 which is multiplied by the noise level (2.5% of the maximum measurement value).

[0129] Fig. 9 is a schematic of an implemented sensor model in Simulink. The original signal is transferred by the response time transfer function. For considering a more realistic situation two transfer functions have been modelled, (i) a response time (T90) of 5 min and system order of n=2 (Eq. 43) represented in time domain in Eq. 44 with r=T90 / 3.89 (ii) T90 as 20 min and system order of n=8 (Eq. 45), with T=T90 / 11.77 . For a better display of the sensor model, the performance of the sensor model against step change input is shown in figure 9. While the ratio of the delay time to rise time (T10 / T90) for the 2nd is low (0.133) this ratio is higher for the 8th order system (0.39)Control Handle

[0130] There is a time gap from the time signal reaches the sensor until the implementation of this signal by the actuator. To have more realistic modelling, it is preferred to implement this response time. Fig. 10 shows the actuator model. Response time for signal implementation has been considered as a 2nd order function. System users can also device for considering an ideal actuator or considering actuator with response time. The following control handles have been considered for implementation in benchmark simulation:• Inlet flow rate (QI)• Flow rate of purge stream and stream number 4, 9 and 11 (QP, Q4, Q9, Qll)• Rotational speed of the pump• Rotational speed of CentrifugeSystem Framework

[0131] Fig. 11 illustrates a framework for conducting parameter analysis within a biomanufacturing process. This framework offers a systemic method for generating valuable insights utilizing the KTB1 structure and current operational data.

[0132] In the initial step, the overarching objective of utilizing the simulation model is defined. Potential objectives may include process monitoring, economic evaluation, fault detection, and diagnosis, as well as sensitivity and uncertainty analysis. Subsequently, in the second step, all pertinent data about the case study product are systematically compiled to facilitate the analysis. It is recommended for the user to meticulously review the initial constant parameters and variables during this phase.

[0133] Moving to step 3, the gathered information is employed to align the case study data with the benchmark model. Following the initial run, based on the outcomes and considering alternative processes constrained by regulatory agencies, step 4 involves investigating the development of alternative processes to optimize the fundamental process flow diagram of the system. In step 5, once process alternatives are identified and developed, it becomes imperative to evaluate their dynamic performance. Finally, in step 6, the results of process simulation are utilized to investigate the research goal and provide a comprehensive analysis.Quantum Algorithms

[0134] Quantum-inspired algorithms for search, optimization, and sampling are important for a domain like bioprocess engineering where the dimensionality of the search space is very high. These algorithms also aid in the representation of complex data structures more naturally and compactly than classical systems, which can make certain machine learning tasks easier. However, quantum algorithms often have high computational and hardware requirements. On the flip side, a completely classical route requires tremendous compute for large-scale monitoring. This would specifically be challenging when a computer is also used for counterfactual simulations. Accordingly, some embodiments of the present disclosure combine hard multifaceted data with soft sensing using quantum computing, generative modeling, and predictions using inline sensors. Within some embodiments, certain problems are solved using quantum computing and others use classical computing (Fig. 12A-B). The hybrid approach may allow for speed-ups in metabolic pathway estimation using quantum semidefinite programming and quadratic unconstrained binary optimization (QUBO),mechanism identification and compressed sensing, and estimation of process variables and cell state via generative Al, like diffusive models.Combining Synthetic Data Generation and Quantum Optimization

[0135] Scale-up and monitoring challenges make the transition from lab-scale biomolecule development to bioreactor production to bring the product to society, demanding. There is a need to understand how microorganisms’ genomic-scale metabolic model affects and is affected by the cultivation media which in turn is affected by bioreactor operation inputs like agitation, aeration, substrate feeding, etc.

[0136] Some embodiments of the present disclosure recognize the symbiotic effect of using soft sensors and synthetic data generation with quantum optimization. In these embodiments, synthetic data and soft sensors play a similar role in the process of either augmenting the input to a model or expanding the output of a model. The model can be based on mathematical correlations, such as an artificial neural network, or probabilistic modelling for distributions. The input to the model may be one or more process conditions corresponding to inline sensor data and distribution information from scaled-down processes. The output can include, but is not limited to:1. Metabolite concentration;2. Enzyme activity;3. Intracellular parameters;4. Culture viability and morphology;5. Genetic expression levels;6. Details biomass content;7. Concentration of trace elements;8. Heterogeneity and distribution of biomass inside the reactor; and / or9. Diffusion coefficients and nutrient solubility in fermentation media.

[0137] In one aspect, synthetic data can be generated to help infer or train mathematical models based on fermentation data. To build useful models, one needs to have sufficient data from experiments to inform the model generation. As the quantity of data may be limited by the number and cost of experiments, synthetic data and / or soft sensors may be used to provide sufficient support to construct the model. Once constructed, one may explore the model using classical or quantum optimization methods to discover new in the optima in thebiomanufacturing design. The synthetic data provides a richer dataset that covers a broader spectrum of scenarios or conditions; this is particularly valuable in optimization problems where the quality and diversity of training data directly impact the model's accuracy and generalizability. In biomanufacturing, the combined approach can optimize fermentation conditions — such as temperature, pH, nutrient supply, and oxygen levels — to maximize yield or product quality. In particular, we expect quantum optimization to quickly explore the complex landscape of possible conditions represented by the synthetic data, identifying optimal settings more efficiently than classical optimization techniques. The process can be iterative, with insights gained from quantum optimization fed back to refine the synthetic data generation process. This feedback loop can lead to continually improved models and optimization outcomes.

[0138] In another aspect, synthetic data and virtual sensors can be used post-optimization.For example, after optimizing a process or system, synthetic data can be used to test the robustness and reliability of the optimization under a wide range of scenarios, including extreme or rare conditions not covered by the original dataset. This can help identify potential weaknesses or areas for further improvement. Alternatively, soft sensors can enhance the result of the model optimization by providing additional data points for measurements that are otherwise impossible or impractical to measure. For example, in Veloso et al. "Monitoring of fed-batch E. coli fermentations with softwaresensors." Bioprocess and biosystems engineering 32 (2009): 381-388, Asymptotic Observer (AO) was applied for the monitoring of fed-batch E. coli fermentation using on-line measurements of common parameters such as dissolved oxygen and carbon dioxide concentrations and the gaseous transfer rates. With AO, Veloso et al. showed that soft sensors can be used for reconstructing the missing state variables even when the process is not exponentially observable, and the kinetics are unknown. This can be particularly useful for predicting the performance of optimized systems under future or hypothetical conditions.

[0139] An exemplary model combining soft sensors and generative Al with quantum optimization may have an input similar to that of the model used to generate synthetic fermentation measurements and the soft sensors. Outputs may include optimized operating conditions. The model itself can be any suitable model previously discussed, with either a quantum or classical optimization framework.

[0140] In some embodiments, a QCAI-TUBE or Q\| / -TUBE is a data-agnostic, validated, modular soft-sensor probe, that uses quantum computing and generative artificial intelligence to provide real-time monitoring solutions for a bioreactor. It also allows for the generation of counterfactual simulations from reactor inputs. Q\| / -TUBE may be a cloud solution for biomanufacturing, monitoring, and scale-up.

[0141] Figs. 12A-B illustrate an embodiment of the Q'P-Tube. In Fig. 12A, the Q -Tube combines a MicroTUBE and a MacroTUBE module. In Fig. 12B, an illustration of an exemplary quantum computing and Al computer is shown in Q -Tube.

[0142] Q\| / -TUBE combines 2 modules - MicroTUBE for microscale calibration and MacroTUBE for macroscale estimation and optimization, each with experimental and computational parts (Fig. 12A). Two candidates are considered prominently used for novel protein manufacturing: Aspergillus niger2 and Pichia pastoris3. Microscale imaging and spectroscopy is conducted based on known factors / variables that affect cell dynamics, like 02 level, CO2 level, pH and temperature, to model a variable’s effect on the metabolic pathway and cell states. These factors are used as seed values to inform generative models to predict cell morphology and viability. Proteomic analysis is used for quantum search and optimization to identify the metabolic pathway leading to the formation of desirable proteins (Fig. 12B)

[0143] In MacroTUBE, tracer variability experiments are conducted (Fig. 12A) in see- through reactors. Reactors are run at different agitation speeds, aeration velocity and viscosity media to understand the effects of these inputs on mixing. Generative Al creates dynamic tracer profdes with operating conditions as input. The independently optimized modules are integrated into Q\| / -TUBE for monitoring and counterfactual simulations. Many variables that cannot be currently measured in reactors will be realized by Q -TUBE until such sensing technologies, when-and-if do exist, become viable.

[0144] In various embodiments, the methods and systems described herein further comprise an anomaly-detection system configured to autonomously monitor the manufacturing process and identify deviations from expected operational behavior. As biomanufacturing and chemical manufacturing processes scale, they become increasingly susceptible to multifaceted disturbances, such as sensor drift, biological contamination, or mechanical degradation. To address this, the implemented anomaly-detection system leverages anarchitecture comprising at least one generator and at least one discriminator. By integrating this anomaly-detection system with the previously described soft sensors and synthetic data generation modules, the overall control system achieves a highly responsive, autonomous optimization loop capable of distinguishing between transient process noise and genuine systemic anomalies.

[0145] In various embodiments, the anomaly-detection system may comprise diverse neural network architectures tailored to the specific dynamics of the manufacturing process. The at least one generator and the at least one discriminator may be implemented utilizing Generative Adversarial Networks (GANs) or Associative Adversarial Networks (AANs). Additionally, autoencoders may be employed to compress and reconstruct operational data to establish a baseline of normal operations. Depending on the specific nature of the at least one measurement, these networks may incorporate sequential deep-learning models, such as Long Short-Term Memory (LSTM) networks or Gated Recurrent Units (GRUs). Sequential models may capture the time-dependent behaviors and complex temporal dependencies inherent in batch or continuous manufacturing processes. For spatially structured data or complex multivariate inputs, Convolutional Neural Networks (CNNs) may also be integrated to extract hierarchical features from data.

[0146] According to certain aspects, at least one of the generator and the discriminator is trained on an operation dataset. The operation dataset may, for example, be supervised, unsupervised, or semi-supervised, or a combination of any such labeling methods. The operation dataset may comprise vast quantities of historical, real-time, and synthetic data representing the "normal" or baseline behavior of the manufacturing system under optimal or acceptable conditions. Because anomalies in complex manufacturing environments are often rare, unpredictable, or entirely novel, supervised learning approaches relying on labeled anomaly data are frequently insufficient. Instead, the anomaly-detection system utilizes unsupervised learning to map the multi-dimensional manifold of normal system behavior. Anomalies (whether singular, isolated events or a plurality of compounding deviations) are subsequently detected based on their statistical or structural divergence from this learned normal behavior.

[0147] In some embodiments, the neural networks (e.g., the generator and discriminator) may be configured to minimize false positive anomaly alerts. False positives are a significantchallenge in large-scale manufacturing, often leading to unnecessary process shutdowns or operator fatigue. To mitigate this, the generator and discriminator may work in concert within an adversarial or reconstructive framework. For example, when an autoencoder or generator is trained to reconstruct the normal operational data, real-time measurements are fed into the system, and the generator attempts to reconstruct the data based on its learned understanding of normal behavior. The discriminator then evaluates the difference (e.g., the reconstruction error) between the actual real-time measurements and the generator's reconstruction. By utilizing an adversarial training process, the discriminator can identify true anomalies while the generator becomes robust at modeling the variance of normal operations, thereby reducing the rate of false positives.

[0148] The architecture of the anomaly-detection system may be implemented using various computational modalities, including purely classical, purely quantum, or hybrid classical- quantum frameworks, depending on the complexity of the process and available computational resources. In a classical-classical embodiment, both the data distributions and the machine learning models (the generator and discriminator) are executed on classical computing hardware. This approach utilizes deep learning frameworks to process the operation dataset and execute the anomaly detection.

[0149] In certain hybrid embodiments, the anomaly-detection system may partition tasks between classical and quantum resources. For instance, in one embodiment, quantum hardware or quantum-inspired tensor networks may be utilized to model the probability distributions of the underlying manufacturing process. The outputs of these quantum distributions are then fed into classical neural networks, which serve as the generator and discriminator to perform the final classification and anomaly scoring. This enables the system to capture process correlations that classical systems might miss, while leveraging the speed and reliability of classical hardware for the adversarial machine learning tasks.

[0150] Alternatively, classical systems may be used to manage, store, and provide the operation dataset and its underlying distributions, while the machine learning components (the generator and / or discriminator) are implemented using quantum computing architectures. In this configuration, quantum neural networks or variational quantum circuits are trained to discriminate between normal and anomalous data; the quantum discriminator can evaluate the classical distributions across a vastly expanded parameter space, enablingthe detection of subtle, multi-variable anomalies that would be computationally intractable for classical neural networks to identify.

[0151] To train the anomaly-detection system, certain embodiments involve generating a benchmark dataset of time-series process data. This dataset may be produced by utilizing a high-fidelity dynamic simulation model of an integrated manufacturing system, which maps both upstream processes (such as cell cultivation and fermentation) and downstream processing operations (such as centrifugation, filtration, and chromatography). The simulation model may be configured to output time-series data spanning both normal operating regimes and anomalous operating regimes. By simulating these conditions in silico, the system can synthesize a high-volume dataset that captures the complex, non-linear dynamics and spatiotemporal gradients of the manufacturing process without requiring costly, risky, or disruptive physical experiments.

[0152] The system may train an ensemble of Generative Adversarial Networks (GANs). In some embodiments, this ensemble may be trained exclusively on the time-series data corresponding to the normal operating regime. Each GAN within the ensemble comprises a generator neural network and a discriminator neural network. During training, the generator may receive a latent vector sampled specifically from a quantum probability distribution. Utilizing quantum hardware or quantum-inspired tensor networks to sample this latent space allows the generator to access a highly complex, multidimensional parameter space that classical probability distributions cannot efficiently replicate. The generator may map this quantum-sampled latent vector to synthetic process data, while the discriminator is trained to distinguish between this generated synthetic data and the actual normal process data from the benchmark dataset. Because the ensemble is exposed solely to normal operational data, the discriminators become hyper-specialized in recognizing the precise statistical boundaries of optimal system behavior.

[0153] The trained discriminators of the GAN ensemble may be isolated and stored for deployment within the active control architecture. During the real-time operational phase, the ensemble of discriminators may be applied continuously to an incoming time-series data stream acquired from the live manufacturing process. Each discriminator can independently process the live input stream and compute an anomaly score representing the input's deviation from its uniquely learned distribution of normal operation. The anomaly-detectionsystem monitors these concurrent evaluations and is configured to label the input stream as anomalous if at least one discriminator in the ensemble scores the stream outside the established bounds of normal operation. This multi-discriminator, ensemble-based scoring mechanism can provide sensitivity to subtle, multivariate process deviations, allowing the system to rapidly identify novel fault conditions and proactively optimize the manufacturing environment.Quantum Computing

[0154] Various physical embodiments of a quantum computer are suitable for use according to the present disclosure. In general, the fundamental data storage unit in quantum computing is the quantum bit, or qubit. The qubit is a quantum-computing analog of a classical digital computer system bit. A classical bit is considered to occupy, at any given point in time, one of two possible states corresponding to the binary digits (bits) 0 or 1. By contrast, a qubit is implemented in hardware by a physical medium with quantum-mechanical characteristics. Such a medium, which physically instantiates a qubit, may be referred to herein as a “physical instantiation of a qubit,” a “physical embodiment of a qubit,” a “medium embodying a qubit,” or similar terms, or simply as a “qubit,” for ease of explanation. It should be understood, therefore, that references herein to “qubits” within descriptions of embodiments described herein refer to physical media which embody qubits.

[0155] Each qubit has an infinite number of different potential quantum-mechanical states.When the state of a qubit is physically measured, the measurement produces one of two different basis states resolved from the state of the qubit. Thus, a single qubit can represent a one, a zero, or any quantum superposition of those two qubit states; a pair of qubits can be in any quantum superposition of 4 orthogonal basis states; and three qubits can be in any superposition of 8 orthogonal basis states. The function that defines the quantum-mechanical states of a qubit is known as its wavefunction. The wavefunction also specifies the probability distribution of outcomes for a given measurement. A qubit, which has a quantum state of dimension two (i.e., has two orthogonal basis states), may be generalized to a d- dimensional “qudit,” where d may be any integral value, such as 2, 3, 4, or higher. In the general case of a qudit, measurement of the qudit produces one of d different basis states resolved from the state of the qudit. Any reference herein to a qubit should be understood to refer more generally to a d-dimensional qudit with any value of d.

[0156] Although certain descriptions of qubits herein may describe such qubits in terms of their mathematical properties, each such qubit may be implemented in a physical medium in any of a variety of different ways. Examples of such physical media include superconducting material, trapped ions, photons, optical cavities, individual electrons trapped within quantum dots, point defects in solids (e.g., phosphorus donors in silicon or nitrogen-vacancy centers in diamond), molecules (e.g., alanine, vanadium complexes), or aggregations of any of the foregoing that exhibit qubit behavior, that is, comprising quantum states and transitions therebetween that can be controllably induced or detected.

[0157] For any given medium that implements a qubit, any of a variety of properties of that medium may be chosen to implement the qubit. For example, if electrons are chosen to implement qubits, then the x component of its spin degree of freedom may be chosen as the property of such electrons to represent the states of such qubits. Alternatively, the y component, or the z component of the spin degree of freedom may be chosen as the property of such electrons to represent the state of such qubits. This is merely a specific example of the general feature that for any physical medium that is chosen to implement qubits, there may be multiple physical degrees of freedom (e.g., the x, y, and z components in the electron spin example) that may be chosen to represent 0 and 1. For any particular degree of freedom, the physical medium may controllably be put in a state of superposition, and measurements may then be taken in the chosen degree of freedom to obtain readouts of qubit values.

[0158] Certain implementations of quantum computers, referred to as gate model quantum computers, comprise quantum gates. In contrast to classical gates, there is an infinite number of possible single-qubit quantum gates that change the state vector of a qubit. Changing the state of a qubit state vector typically is referred to as a single-qubit rotation, and may also be referred to herein as a state change or a single-qubit quantum-gate operation. A rotation, state change, or single-qubit quantum-gate operation may be represented mathematically by a unitary 2X2 matrix with complex elements. A rotation corresponds to a rotation of a qubit state within its Hilbert space, which may be conceptualized as a rotation of the Bloch sphere. (As is well-known to those having ordinary skill in the art, the Bloch sphere is a geometrical representation of the space of pure states of a qubit.) Multi-qubit gates alter the quantum state of a set of qubits. For example, two-qubit gates rotate the state of two qubits as a rotation in the four-dimensional Hilbert space of the two qubits. (As is well-known to those havingordinary skill in the art, a Hilbert space is an abstract vector space possessing the structure of an inner product that allows length and angle to be measured. Furthermore, Hilbert spaces are complete: there are enough limits in the space to allow the techniques of calculus to be used.)

[0159] A quantum circuit may be specified as a sequence of quantum gates. As described in more detail below, the term “quantum gate,” as used herein, refers to the application of a gate control signal (defined below) to one or more qubits to cause those qubits to undergo certain physical transformations and thereby to implement a logical gate operation. To conceptualize a quantum circuit, the matrices corresponding to the component quantum gates may be multiplied together in the order specified by the gate sequence to produce a 2nX2n complex matrix representing the same overall state change on n qubits. A quantum circuit may thus be expressed as a single resultant operator. However, designing a quantum circuit in terms of constituent gates allows the design to conform to a standard set of gates, and thus enable greater ease of deployment. A quantum circuit thus corresponds to a design for actions taken upon the physical components of a quantum computer.

[0160] A given variational quantum circuit may be parameterized in a suitable devicespecific manner. More generally, the quantum gates making up a quantum circuit may have an associated plurality of tuning parameters. For example, in embodiments based on optical switching, tuning parameters may correspond to the angles of individual optical elements.

[0161] In certain embodiments of quantum circuits, the quantum circuit includes both one or more gates and one or more measurement operations. Quantum computers implemented using such quantum circuits are referred to herein as implementing “measurement feedback.” For example, a quantum computer implementing measurement feedback may execute the gates in a quantum circuit and then measure only a subset (i.e., fewer than all) of the qubits in the quantum computer, and then decide which gate(s) to execute next based on the outcome(s) of the measurement(s). In particular, the measurement(s) may indicate a degree of error in the gate operation(s), and the quantum computer may decide which gate(s) to execute next based on the degree of error. The quantum computer may then execute the gate(s) indicated by the decision. This process of executing gates, measuring a subset of the qubits, and then deciding which gate(s) to execute next may be repeated any number of times.Measurement feedback may be useful for performing quantum error correction, but is not limited to use in performing quantum error correction. For every quantum circuit, there is anerror-corrected implementation of the circuit with or without measurement feedback.

[0162] Some embodiments described herein generate, measure, or utilize quantum states that approximate a target quantum state (e.g., a ground state of a Hamiltonian). As will be appreciated by those trained in the art, there are many ways to quantify how well a first quantum state “approximates” a second quantum state. In the following description, any concept or definition of approximation known in the art may be used without departing from the scope hereof. For example, when the first and second quantum states are represented as first and second vectors, respectively, the first quantum state approximates the second quantum state when an inner product between the first and second vectors (called the “fidelity” between the two quantum states) is greater than a predefined amount (typically labeled c). In this example, the fidelity quantifies how “close” or “similar” the first and second quantum states are to each other. The fidelity represents a probability that a measurement of the first quantum state will give the same result as if the measurement were performed on the second quantum state. Proximity between quantum states can also be quantified with a distance measure, such as a Euclidean norm, a Hamming distance, or another type of norm known in the art. Proximity between quantum states can also be defined in computational terms. For example, the first quantum state approximates the second quantum state when a polynomial time-sampling of the first quantum state gives some desired information or property that it shares with the second quantum state.

[0163] Not all quantum computers are gate model quantum computers. Embodiments of the techniques described herein are not limited to being implemented using gate model quantum computers. As an alternative example, embodiments of the techniques described herein may be implemented, in whole or in part, using a quantum computer that is implemented using a quantum annealing architecture, which is an alternative to the gate model quantum computing architecture. More specifically, quantum annealing (QA) is a metaheuristic for finding the global minimum of a given objective function over a given set of candidate solutions (candidate states), by a process using quantum fluctuations.

[0164] Fig. 14B shows a diagram illustrating operations typically performed by a computer system 250 which implements quantum annealing. The computer system 250 includes both a quantum computer 252 and a classical computer 254. Operations shown on the left of the dashed vertical line 256 typically are performed by the quantum computer 252, whileoperations shown on the right of the dashed vertical line 256 typically are performed by the classical computer 254.

[0165] Quantum annealing starts with the classical computer 254 generating an initial Hamiltonian 260 and a final Hamiltonian 262 based on a computational problem 258 to be solved, and providing the initial Hamiltonian 260, the final Hamiltonian 262 and an annealing schedule 270 as input to the quantum computer 252. The quantum computer 252 prepares an initial state 266 (FIG. 14B, operation 264), such as a quantum-mechanical superposition of all possible states (candidate states) with equal weights, based on the initial Hamiltonian 260. The classical computer 254 provides the initial Hamiltonian 260, a final Hamiltonian 262, and an annealing schedule 270 to the quantum computer 252. The quantum computer 252 starts in the initial state 266, and evolves its state according to the annealing schedule 270 following the time-dependent Schrodinger equation, a natural quantummechanical evolution of physical systems (Fig. 14B, operation 268). More specifically, the state of the quantum computer 252 undergoes time evolution under a time-dependent Hamiltonian, which starts from the initial Hamiltonian 260 and terminates at the final Hamiltonian 262. If the rate of change of the system Hamiltonian is slow enough, the system stays close to the ground state of the instantaneous Hamiltonian. If the rate of change of the system Hamiltonian is accelerated, the system may leave the ground state temporarily but produce a higher likelihood of concluding in the ground state of the final problem Hamiltonian, i.e., diabatic quantum computation. At the end of the time evolution, the set of qubits on the quantum annealer is in a final state 272, which is expected to be close to the ground state of the classical Ising model that corresponds to the solution to the computational problem 258. An experimental demonstration of the success of quantum annealing for random magnets was reported immediately after the initial theoretical proposal.

[0166] The final state 272 of the quantum computer 252 is measured, thereby producing results 276 (i.e., measurements) (Fig. 14B, measurement operation 274). The measurement operation 274 may be performed, for example, in any of the ways disclosed herein, such as in any of the ways disclosed herein in connection with the measurement unit 110 in Fig. 13. The classical computer 254 performs postprocessing on the results 276 to produce output 280 representing a solution to the computational problem 258 (Fig. 14B, operation 278).

[0167] As yet another alternative example, embodiments of the techniques described hereinmay be implemented, in whole or in part, using a quantum computer that is implemented using a one-way quantum computing architecture, also referred to as a measurement-based quantum computing architecture, which is another alternative to the gate model quantum computing architecture. More specifically, the one-way or measurement-based quantum computer (MBQC) is a method of quantum computing that first prepares an entangled resource state, usually a cluster state or graph state, then performs single qubit measurements on it. It is "one-way" because the resource state is destroyed by the measurements.

[0168] The outcome of each individual measurement is random, but they are related in such a way that the computation always succeeds. In general, the choices of basis for later measurements need to depend on the results of earlier measurements, and hence the measurements cannot all be performed at the same time.

[0169] Any of the functions disclosed herein may be implemented using means for performing those functions. Such means include, but are not limited to, any of the components disclosed herein, such as the computer-related components described below.

[0170] Referring to Fig. 13, a diagram is shown of a system 100 implemented according to one example embodiment. Referring to FIG. 14A, a flowchart is shown of a method 200 performed by the system 100 of Fig. 13 according to one example embodiment. The system 100 includes a quantum computer 102. The quantum computer 102 includes a plurality of qubits 104, which may be implemented in any of the ways disclosed herein. There may be any number of qubits 104 in the quantum computer 102. For example, the qubits 104 may include or consist of no more than 2 qubits, no more than 4 qubits, no more than 8 qubits, no more than 16 qubits, no more than 32 qubits, no more than 64 qubits, no more than 128 qubits, no more than 256 qubits, no more than 512 qubits, no more than 1024 qubits, no more than 2048 qubits, no more than 4096 qubits, or no more than 8192 qubits. These are merely examples, in practice there may be any number of qubits 104 in the quantum computer 102.

[0171] There may be any number of gates in a quantum circuit. However, in some embodiments the number of gates may be at least proportional to the number of qubits 104 in the quantum computer 102. In some embodiments the gate depth may be no greater than the number of qubits 104 in the quantum computer 102, or no greater than some linear multiple of the number of qubits 104 in the quantum computer 102 (e.g., 2, 3, 4, 5, 6, or 7).

[0172] The qubits 104 may be interconnected in any graph pattern. For example, they beconnected in a linear chain, a two-dimensional grid, an all-to-all connection, any combination thereof, or any subgraph of any of the preceding.

[0173] As will become clear from the description below, although the quantum computer 102 is referred to herein as a “quantum computer,” this does not imply that all components of the quantum computer 102 leverage quantum phenomena. One or more components of the quantum computer 102 may, for example, be classical (i.e., non-quantum components) components which do not leverage quantum phenomena.

[0174] The quantum computer 102 includes a control unit 106, which may include any of a variety of circuitry and / or other machinery for performing the functions disclosed herein. The control unit 106 may, for example, consist entirely of classical components. The control unit 106 generates and provides as output one or more control signals 108 to the qubits 104. The control signals 108 may take any of a variety of forms, such as any kind of electromagnetic signals, such as electrical signals, magnetic signals, optical signals (e.g., laser pulses), or any combination thereof.

[0175] In embodiments in which some or all of the qubits 104 are implemented as photons (also referred to as a “quantum optical” implementation) that travel along waveguides, the control unit 106 may be a beam splitter (e.g., a heater or a mirror), the control signals 108 may be signals that control the heater or the rotation of the mirror, the measurement unit 110 may be a photodetector, and the measurement signals 112 may be photons.

[0176] In embodiments in which some or all of the qubits 104 are implemented as charge type qubits (e.g., transmon, X-mon, G-mon) or flux-type qubits (e.g., flux qubits, capacitively shunted flux qubits) (also referred to as a “circuit quantum electrodynamic” (circuit QED) implementation), the control unit 106 may be a bus resonator activated by a drive, the control signals 108 may be cavity modes, the measurement unit 110 may be a second resonator (e.g., a low-Q resonator), and the measurement signals 112 may be voltages measured from the second resonator using dispersive readout techniques.

[0177] In embodiments in which some or all of the qubits 104 are implemented as superconducting circuits, the control unit 106 may be a circuit QED-assisted control unit or a direct capacitive coupling control unit or an inductive capacitive coupling control unit, the control signals 108 may be cavity modes, the measurement unit 110 may be a second resonator (e.g., a low-Q resonator), and the measurement signals 112 may be voltagesmeasured from the second resonator using dispersive readout techniques.

[0178] In embodiments in which some or all of the qubits 104 are implemented as trapped ions (e.g., electronic states of, e.g., magnesium ions), the control unit 106 may be a laser, the control signals 108 may be laser pulses, the measurement unit 110 may be a laser and either a CCD or a photodetector (e.g., a photomultiplier tube), and the measurement signals 112 may be photons.

[0179] In embodiments in which some or all of the qubits 104 are implemented using nuclear magnetic resonance (NMR) (in which case the qubits may be molecules, e.g., in liquid or solid form), the control unit 106 may be a radio frequency (RF) antenna, the control signals 108 may be RF fields emitted by the RF antenna, the measurement unit 110 may be another RF antenna, and the measurement signals 112 may be RF fields measured by the second RF antenna.

[0180] In embodiments in which some or all of the qubits 104 are implemented as nitrogenvacancy centers (NV centers), the control unit 106 may, for example, be a laser, a microwave antenna, or a coil, the control signals 108 may be visible light, a microwave signal, or a constant electromagnetic field, the measurement unit 110 may be a photodetector, and the measurement signals 112 may be photons.

[0181] In embodiments in which some or all of the qubits 104 are implemented as two- dimensional quasiparticles called “anyons” (also referred to as a “topological quantum computer” implementation), the control unit 106 may be nanowires, the control signals 108 may be local electrical fields or microwave pulses, the measurement unit 110 may be superconducting circuits, and the measurement signals 112 may be voltages.

[0182] In embodiments in which some or all of the qubits 104 are implemented as semiconducting material (e.g., nanowires), the control unit 106 may be microfabricated gates, the control signals 108 may be RF or microwave signals, the measurement unit 110 may be microfabricated gates, and the measurement signals 112 may be RF or microwave signals.

[0183] Although not shown explicitly in Fig. 13 and not required, the measurement unit 110 may provide one or more feedback signals 114 to the control unit 106 based on the measurement signals 112. For example, quantum computers referred to as “one-way quantum computers” or “measurement-based quantum computers” utilize the feedback signals 114 from the measurement unit 110 to the control unit 106. The feedback signals 114 are alsonecessary for the operation of fault-tolerant quantum computing and error correction.

[0184] The control signals 108 may, for example, include one or more state preparation signals which, when received by the qubits 104, cause some or all of the qubits 104 to change their states. Such state preparation signals constitute a quantum circuit also referred to as an “ansatz circuit.” The resulting state of the qubits 104 is referred to herein as an “initial state” or an “ansatz state.” The process of outputting the state preparation signal(s) to cause the qubits 104 to be in their initial state is referred to herein as “state preparation” (FIG. 14A, section 206). A special case of state preparation is “initialization,” also referred to as a “reset operation,” in which the initial state is one in which some or all of the qubits 104 are in the “zero” state i.e. the default single-qubit state. More generally, state preparation may involve using the state preparation signals to cause some or all of the qubits 104 to be in any distribution of desired states. In some embodiments, the control unit 106 may first perform initialization on the qubits 104 and then perform preparation on the qubits 104, by first outputting a first set of state preparation signals to initialize the qubits 104, and by then outputting a second set of state preparation signals to put the qubits 104 partially or entirely into non-zero states.

[0185] Another example of control signals 108 that may be output by the control unit 106 and received by the qubits 104 are gate control signals. The control unit 106 may output such gate control signals, thereby applying one or more gates to the qubits 104. Applying a gate to one or more qubits causes the set of qubits to undergo a physical state change which embodies a corresponding logical gate operation (e.g., single-qubit rotation, two-qubit entangling gate or multi-qubit operation) specified by the received gate control signal. As this implies, in response to receiving the gate control signals, the qubits 104 undergo physical transformations which cause the qubits 104 to change state in such a way that the states of the qubits 104, when measured (see below), represent the results of performing logical gate operations specified by the gate control signals. The term “quantum gate,” as used herein, refers to the application of a gate control signal to one or more qubits to cause those qubits to undergo the physical transformations described above and thereby to implement a logical gate operation.

[0186] It should be understood that the dividing line between state preparation (and the corresponding state preparation signals) and the application of gates (and the correspondinggate control signals) may be chosen arbitrarily. For example, some or all the components and operations that are illustrated in Figs. 13 and 14A-2B as elements of “state preparation” may instead be characterized as elements of gate application. Conversely, for example, some or all of the components and operations that are illustrated in Figs. 13 and 14A-2B as elements of “gate application” may instead be characterized as elements of state preparation. As one particular example, the system and method of Figs. 13 and 14A-2B may be characterized as solely performing state preparation followed by measurement, without any gate application, where the elements that are described herein as being part of gate application are instead considered to be part of state preparation. Conversely, for example, the system and method of Figs. 13 and 14A-2B may be characterized as solely performing gate application followed by measurement, without any state preparation, and where the elements that are described herein as being part of state preparation are instead considered to be part of gate application.

[0187] Fig. 15 is a schematic of a quantum and classical computer, in accordance with one or more embodiments of this disclosure.

[0188] The quantum computer 102 also includes a measurement unit 110, which performs one or more measurement operations on the qubits 104 to read out measurement signals 112 (also referred to herein as “measurement results”) from the qubits 104, where the measurement signals 112 are signals representing the states of some or all of the qubits 104. In practice, the control unit 106 and the measurement unit 110 may be entirely distinct from each other, or contain some components in common with each other, or be implemented using a single unit (i.e., a single unit may implement both the control unit 106 and the measurement unit 110). For example, a laser unit may be used both to generate the control signals 108 and to provide stimulus (e.g., one or more laser beams) to the qubits 104 to cause the measurement signals 112 to be generated.

[0189] In general, the quantum computer 102 may perform various operations described above any number of times. For example, the control unit 106 may generate one or more control signals 108, thereby causing the qubits 104 to perform one or more quantum gate operations. The measurement unit 110 may then perform one or more measurement operations on the qubits 104 to read out a set of one or more measurement signals 112. The measurement unit 110 may repeat such measurement operations on the qubits 104 before the control unit 106 generates additional control signals 108, thereby causing the measurementunit 110 to read out additional measurement signals 112 resulting from the same gate operations that were performed before reading out the previous measurement signals 112. The measurement unit 110 may repeat this process any number of times to generate any number of measurement signals 112 corresponding to the same gate operations. The quantum computer 102 may then aggregate such multiple measurements of the same gate operations in any of a variety of ways.

[0190] After the measurement unit 110 has performed one or more measurement operations on the qubits 104 after they have performed one set of gate operations, the control unit 106 may generate one or more additional control signals 108, which may differ from the previous control signals 108, thereby causing the qubits 104 to perform one or more additional quantum gate operations, which may differ from the previous set of quantum gate operations. The process described above may then be repeated, with the measurement unit 110 performing one or more measurement operations on the qubits 104 in their new states (resulting from the most recently performed gate operations).

[0191] In general, the system 100 may implement a plurality of quantum circuits as follows.For each quantum circuit C in the plurality of quantum circuits (Fig. 14A, operation 202), the system 100 performs a plurality of “shots” on the qubits 104. The meaning of a shot will become clear from the description that follows. For each shot S in the plurality of shots (Fig.14A, operation 204), the system 100 prepares the state of the qubits 104 (Fig. 14A, section 206). More specifically, for each quantum gate G in quantum circuit C (Fig. 14A, operation 210), the system 100 applies quantum gate G to the qubits 104 (Fig. 14A, operations 212 and 214)

[0192] Then, for each of the qubits 104 (Fig. 14A, operation 216), the system 100 measures the qubit 104 to produce measurement output representing a current state of qubit 104 (FIG.14 A, operations 218 and 220).

[0193] The operations described above are repeated for each shot S (Fig. 14A, operation 222), and circuit C (Fig. 14A, operation 224). As the description above implies, a single “shot” involves preparing the state of the qubits 104 and applying all of the quantum gates in a circuit to the qubits 104 and then measuring the states of the qubits 104; and the system 100 may perform multiple shots for one or more circuits.

[0194] The present disclosure may be embodied as a system, a method, and / or a computerprogram product. The computer program product may include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present disclosure.

[0195] The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium may be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. Anon-exhaustive list of more specific examples of the computer readable storage medium includes the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD- ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or raised structures in a groove having instructions recorded thereon, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.

[0196] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network may comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0197] Computer readable program instructions for carrying out operations of the present disclosure may be assembler instructions, instruction-set-architecture (ISA) instructions,machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like, and conventional procedural programming languages, such as the “C” programming language or similar programming languages. The computer readable program instructions may execute entirely on the user’s computer, partly on the user’s computer, as a stand-alone software package, partly on the user’s computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user’s computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate arrays (FPGA), or programmable logic arrays (PLA) may execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present disclosure.

[0198] Aspects of the present disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer readable program instructions.

[0199] These computer readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer readable program instructions may also be stored in a computer readable storage medium that can direct a computer, a programmable data processing apparatus, and / or other devices to function in a particular manner, such that the computer readable storage medium having instructions stored therein comprises an article ofmanufacture including instructions which implement aspects of the function / act specified in the flowchart and / or block diagram block or blocks.

[0200] The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process, such that the instructions which execute on the computer, other programmable apparatus, or other device implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0201] The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flowchart illustration, and combinations of blocks in the block diagrams and / or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts or carry out combinations of special purpose hardware and computer instructions.

[0202] The descriptions of the various embodiments of the present disclosure have been presented for purposes of illustration, but are not intended to be exhaustive or limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein was chosen to best explain the principles of the embodiments, the practical application or technical improvement over technologies found in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.

[0203] While the disclosure has been described in connection with certain embodiments, it is to be understood that the disclosure is not to be limited to the disclosed embodiments but, onthe contrary, is intended to cover various modifications and equivalent arrangements included within the scope of the appended claims, which scope is to be accorded the broadest interpretation so as to encompass all such modifications and equivalent structures as is permitted under the law.

Claims

1. What is claimed is:

1. A method for autonomous optimization of a manufacturing process, the method comprising:receiving at least one measurement at a production volume for the manufacturing process;applying a generative artificial intelligence (AT) system to generate synthetic data based on the at least one measurement;integrating at least one process model based on the synthetic data to generate a soft sensor that predicts a target measurement in the manufacturing process; andimplementing an anomaly-detection system wherein the anomaly-detection system comprises at least one generator and at least one discriminator wherein at least one of the at least one generator and at least one discriminator are trained on an operation dataset.

2. The method of claim 1, wherein the manufacturing process is a biomanufacturing process.

3. The method of claim 1, wherein the manufacturing process is a chemical manufacturing process.

4. The method of claim 1, wherein the manufacturing process is a commodity biologies manufacturing process.

5. The method of claim 1, wherein the generative Al system comprises a quantum generative Al model.

6. The method of claim 1, further comprising using the soft sensor to generate target data in the manufacturing process.

7. The method of claim 6, wherein the target data is used to monitor the manufacturing process.

8. The method of claim 6, wherein the target data is used for a control system of the manufacturing process.

9. The method of claim 1, further comprising generating a benchmark dataset of time-series process data comprising normal operating regimes and anomalous operating regimes, wherein the benchmark dataset is produced by a dynamic simulation model of an integrated upstream and downstream system, and wherein the operation dataset is derived from the normal operating regimes of the benchmark dataset.

10. The method of claim 9, wherein the anomaly-detection system comprises at least one generative adversarial network comprising:a generator neural network configured to receive a latent vector sampled from a quantum probability distribution and map the latent vector to synthetic process data; anda discriminator neural network trained to distinguish the synthetic process data from data associated with the normal operating regime.

11. The method of claim 10, further comprising:storing the trained discriminator neural network of the at least one generative adversarial network;applying input time-series data stream representing measurements from the manufacturing process into the trained discriminator neural network;computing an anomaly score for the trained discriminator neural network; and labeling the input time-series data stream as anomalous when the discriminator neural network scores the input time-series data stream outside a learned distribution of normal operation.

12. A method for autonomous optimization of a manufacturing process, the method comprising:receiving at least one measurement at a production volume for the manufacturing process;applying a generative artificial intelligence (AT) system to generate synthetic data based on the at least one measurement;integrating at least one process model based on the synthetic data to generate a soft sensor that predicts a target measurement in the manufacturing process; andimplementing an agent-based Al system structured as an ensemble of semi -autonomous agents to perform real-time inference and control, wherein the agent-based Al system utilizes the soft sensor and the synthetic data to execute an agent-based decision process for optimizing at least one condition of the manufacturing process.

13. The method of claim 12, wherein the manufacturing process is a biomanufacturing process.

14. The method of claim 12, wherein the manufacturing process is a chemical manufacturing process.

15. The method of claim 12, wherein the manufacturing process is a commodity biologies manufacturing process.

16. The method of claim 12, wherein the generative Al system comprises a quantum generative Al model.

17. The method of claim 12, further comprising using the soft sensor to generate target data in the manufacturing process.

18. The method of claim 17, wherein the target data is used to monitor the manufacturing process.

19. The method of claim 17, wherein the target data is used for a control system of the manufacturing process.

20. A method for quantum optimization of a manufacturing process, the method comprising:receiving at least one measurement at a production volume for the manufacturing process;applying a quantum generative artificial intelligence (Al) system to generate data based on the at least one measurement;integrating one or more process models based on the generated data;predicting a dynamic of the production volume using a soft sensor; andoptimizing a condition in a large-scale manufacturing process via an agent-based Al system, wherein the agent-based Al system comprises an ensemble of semi-autonomous agents that utilize the dynamic to execute an agent-based decision process for real-time autonomous control.