Number Of Batches Research Articles

A Gentle Introduction to Physics-Informed Neural Networks, with Applications in Static Rod and Beam Problems

A modern approach to solving mathematical models involving differential equations, the so-called Physics-Informed Neural Network (PINN), is based on the techniques which include the use of artificial neural networks and the method of fitting the governing differential equations at collocation points. In this paper, training of the PINN with an application of optimization techniques is performed on simple one-dimensional mechanical problems of elasticity, namely rods and beams. Different boundary conditions are considered. Required computer algorithms are implemented using Python programming packages with the intention of creating neural networks. Numerical results are presented, and the efficiency of the proposed technique is investigated through numerical experiments with different numbers of epochs, batches, hidden layers, neurons, and collocation points. The paper provides useful skills for using a PINN for different problems of solid mechanics. The proposed methodology is a continuation of our intention of using PINNs for problems of the theory of elasticity. The objectives are to present simply the main steps of constructing PINN and an implementation of it. A detailed explanation of the Python programming code, based on the scientific software Tensorflow, built in the Keras library and optimizers, may help compose an effective code for complicated models in mechanics. PINNs are proposed in many recent publications to solve complicated direct and inverse problems. It seems to be a promising method that will play a central role in the development of computational mechanics in the near future. Nevertheless, the lack of educational material does not help new users to enter this scientific area. The present contribution describes the method for the solution of elementary rod and beam problems and gives computer codes that may help the reader to understand the method and to apply it to other problems.

Microplastic loads within riverine fishes and macroinvertebrates are not predictable from ecological or morphological characteristics

Microplastics are a relatively new but important form of freshwater contamination that can be ingested by a range of different species, with particle counts thought to be predictable from species ecology and morphology. Here, we report levels of microplastics in a 26 μm-5 mm size range within the macroinvertebrate and fish community of a lowland river (Dorset Stour, SW England), and test the hypothesis that counts are predictable from characteristics such as feeding guild, body length and trophic position. Macroinvertebrates (n = 257, 12 taxa) and fish (n = 418, 9 species) were collected from distinct river reaches by kick sampling and rod and line angling, respectively. Batches of whole macroinvertebrates and individual fish gastrointestinal tracts were digested with 30% hydrogen peroxide before microplastic screening and FTIR polymer confirmation on a particle subset. Particles were found in 40% of pooled macroinvertebrate batches (taxa incidences: 14–75%) and 39% of fishes (species incidences: 29–47%). Dominant particle feature categories were ≤100 μm, blue/green, fragments and fibres identified as various polyolefins. Although particle counts in macroinvertebrates were highest in Ephemeroptera (mean of 0.74 particles per individual), the relationships between particle loads, batch number and guild were all non-significant. In fishes, particle counts were not significantly related to species, stomach structure, feeding guild or body length, with spatial differences also not apparent across the catchment. Individual fish particle counts were similarly not significantly associated with their trophic positions (calculated from bulk δ15N values for a subset of fishes) and parasite load of Pomphorhynchus tereticollis. Correlations between fish and macroinvertebrate particle counts within specific river reaches were also not significant. In entirety, these results indicated although loadings of microplastic particles were relatively consistent within the two communities, they were not predictable from any of their ecological or morphological characteristics.

I-optimal design of split-plot mixture-process variable experiments: A case study on potato crisps

Designed experiments are powerful tools when developing new products or processes. They allow changing the settings in a systematic way such that a minimal experimental effort results in maximal information. In the food industry, performing designed experiments can be challenging because products and processes are often complex and involve many factors of different kinds. Moreover, some of the factors (for instance, the ingredients of a new formulation) may have intrinsic properties that can be measured but cannot be changed. These are referred to as pre-specified factors or covariates. In this paper, we discuss a representative case study and show how these complications often encountered in the food industry are handled. The goal of the case study is to lower the lipid content of potato crisps by means of a split-plot type of designed experiment involving a complex mixture of ingredients, various constraints on the proportions of the mixture ingredients, and a limited number of batches that each come with a set of covariates. We show how our approach provides insight into the most important factors and allows the product quality to be optimized in an efficient way.

Use of Different Hyperparameter Optimization Algorithms in ANN for Predicting the Compressive Strength of Concrete Containing Calcined Clay

Calcined clays as supplementary cementitious materials (SCMs) have the potential to answer the need for SCMs because they have many resources in comparison with other SCMs. Moreover, calcined clays can be combined with limestone, which makes limestone calcined clay cement (LC3). LC3 can reduce the clinker of cement, which leads to the reduction of carbon dioxide pollution. The application of calcined clays as SCMs can make concrete with lower cost and the same compressive strength in comparison with other SCMs such as fly ash. Therefore, in this research, a predictive model has been built by the application of an artificial neural network (ANN), which can encourage the industry to increase the clay application in concrete. On the other hand, for improving the performance of the ANN model, four different optimization algorithms, such as genetic algorithm (GA), Bayesian optimization with a Gaussian process (BO-GP), Bayesian optimization with tree parzen estimator (BO-TPE), and hyperband algorithm, have been applied to optimize the number of neurons, number of hidden layers, activation function, number of batches, and epochs to predict the compressive strength of concrete. The results performance of models showed the superiority of the hyperband algorithm in running time and accuracy.

International Regulatory Collaboration on the Analysis of Nitrosamines in Metformin-Containing Medicines

Recalls of some batches of metformin have occurred due to the detection of N-nitrosodimethylamine (NDMA) in amounts above the acceptable intake (AI) of 96 ng per day. Prior to the recalls, an international regulatory laboratory network had been monitoring drugs for nitrosamine impurities with each laboratory independently developing and validating multiple analytical procedures to detect and measure nitrosamines in metformin drugs used in their jurisdictions. Here, we provide an overview of the analysis of metformin active pharmaceutical ingredients (APIs) and drug products with 1090 samples (875 finished dosage forms (FDFs) and 215 API samples) tested beginning in November of 2019 through July of 2020. Samples were obtained internationally by a variety of approaches, including purchased, received from firms via information requests or selected by regional regulatory authorities (either at wholesalers or during GMP inspections). Only one nitrosamine (NDMA) was detected and was only present in some batches of metformin products. For API samples, 213 out of 215 lots tested had no measurable level of NDMA. For FDF samples tested, the number of batches with NDMA above the AI amount for patient safety was 17.8% (156/875). Based on these data, although the presence of NDMA was of concern, 82.2% of the samples of metformin drug products tested met quality and safety standards for patients. Regulatory agencies continue to collaborate extensively and work with marketing authorization holders to understand root causes of nitrosamine formation and agree on corrective actions to mitigate the presence of NDMA in future metformin batches.Graphical abstract

Development of stable self-nanoemulsifying composition and its nanoemulsions for improved oral delivery of non-oncology drugs against hepatic cancer

Cancer remains a major health concern and second leading cause of mortality worldwide. The highest cancer-related mortalities are observed in low and middle-income countries (LMIC) due to unaffordable cost of medicines and lack of safe and effective chemotherapy. Therefore, the key goal of the present research was to develop affordable, safe, efficacious and stable nanoemulsion (NE) system co-loaded with Tadalafil (TDF) and Ketoconazole (KTZ) for repurposing against hepatic cancer. The batch number 1 of SNEDDS was selected as optimized batch based on dispersion time, viscosity and stability. The NE displayed particle size of 41±5 nm (PDI: 0.189±0.064) and Zeta potential of -27.1 mV. The NE exhibited controlled in-vitro release of both TDF (52.37 ± 3.41%) and (KTZ 52.31 ± 3.88%) in PBS pH 6.8 after 24 h. Moreover, the in-vitro hemolysis study showed moderate hemolytic activity of NE towards blood cells. In addition, the in-vitro cytotoxicity study revealed more cytotoxicity of NE than plain TDF and KTZ against HepG2 cell line after 48 h of incubation. Moreover, the HepG2 cellular uptake of TDF and KTZ was found to be substantially higher from NE when compared to the plain TDF and KTZ. Furthermore, NE was found to be stable for more than 6 months with no phase separation and creaming. Stable, safe, effective and affordable NE was developed which might be used as an alternative strategy in the treatment of hepatic cancer and its similar anticancer effect against other cancers need to be validated. Further, TDF and KTZ combination would a potentials approach over approved chemotherapy against hepatic cancer.

Applying and Comparing LSTM and ARIMA to Predict CO Levels for a Time-Series Measurements in a Port Area

Air pollution is a major problem in the everyday life of citizens, especially air pollution in the transport domain. Ships play a significant role in coastal air pollution, in conjunction with transport mobility in the broader area of ports. As such, ports should be monitored in order to assess air pollution levels and act accordingly. In this paper, we obtain CO values from environmental sensors that were installed in the broader area of the port of Igoumenitsa in Greece. Initially, we analysed the CO values and we have identified some extreme values in the dataset that showed a potential event. Thereafter, we separated the dataset into 6-h intervals and showed that we have an extremely high rise in certain hours. We transformed the dataset to a moving average dataset, with the objective being the reduction of the extremely high values. We utilised a machine-learning algorithm, namely the univariate long short-term memory (LSTM) algorithm to provide the predicted outcome of the time series from the port that has been collected. We performed experiments by using 100, 1000, and 7000 batches of data. We provided results on the model loss and the root-mean-square error as well as the mean absolute error. We showed that with the case with batch number equals to 7000, the LSTM we achieved a good prediction outcome. The proposed method was compared with the ARIMA model and the comparison results prove the merit of the approach.

Evaluation of plastic packaged water quality using health risk indices: A case study of sachet and bottled water in Accra, Ghana

Plastic packaged water is the drinking water of choice for urban populations across Africa but its quality remains questionable in most developing countries. Six hundred (600) packages, consisting of sachet and bottled water, were sampled from two high-end companies in Accra (Ghana) and stored through their shelf lives under an average room temperature of 30 °C. The samples were tested for physicochemical quality and the presence of bacteria and phthalate esters at 2n × 3 periods, where n is the sampled batch number. The data were described and modelled with embedded Bayesian and Machine Learning algorithms in JASP0.16.0.0 and Argo-4.1.3. The results reported lower than regulated levels of electrical conductivity (163.66 μS/cm), alkalinity (39.67 mg/L), and residual chlorine (<0.01 mg/L) while the pH was generally within specification (6.5–7.7). All samples showed progressive biological contamination following the third week (sachet samples) and the sixth week (bottled water) of incubation. Initial samples, including raw water, processed bulk water and packaged water did not present detectable microbial growth. The total microbial load in sachet samples grew at 0.936 cfu/week and 1.006 cfu/week for the bottled samples although the results did not exceed 1000 cfu/L (0-976 cfu/100 mL). Modelled mean probability of infection was 1.196 × 10−4 in 67% of the samples. Raw and processed water samples did not show detectable levels of phthalate contaminants. The mean hazard index calculated on the individual hazard quotients of phthalates was 7.41 × 10−3 ± 8.20 × 10−4, suggesting lower acute risk potential. Mean integrated lifetime cancer risk (ILCR) was determined to be 1.53 × 10−3 ± 1.71 × 10−4 within a range of 2.86 × 10−4 and 7.18 × 10−3. Mean child ILCR was about 70% of adult ILCR and increased from 4.16 × 10−4 to 2.41 × 10−3 for sachet and 4.93 × 10−4 to 7.18 × 10−3 for bottled water. For adult ILCR, sachet water presented 2.86 × 10−4 to 1.65 × 10−3, and 3.38 × 10−4 to 4.93 × 10−3 for bottled water. This study confirmed the presence of phthalates and pathogenic bacteria in the samples, at-risk levels that require mitigation.

Novel two-dimensional off-policy [formula omitted]-learning method for output feedback optimal tracking control of batch process with unknown dynamics

Reinforcement learning (RL) is an artificial intelligence algorithm that can learn adaptive optimal control law online. In view of the fact that the previous control approaches were usually overly dependent on the model parameters of system, and most existing RL methods are based on state feedback, their application in actual industrial production is limited. Additionally, developing accurate process system models and ensuring the closed-loop system’s control performance is more challenging, as modern businesses place a premium on product quality and economic efficiency. As a result, this work introduces a novel data-driven two-dimensional (2D) off-policy Q-learning method based on output feedback is used to achieve optimal tracking control for batch process. First, the error between the actual output and the given set-point is extended to the system to ensure the good tracking performance. Second, by analyzing the relationship between the value function and the Q-function obtained from the 2D system’s performance index, the 2D Bellman equation is obtained in terms of output feedback that is independent of the model parameters. The optimal control problem can be effectively solved by the proposed method in this paper when the policy iteration is executed using only the measurement data of system along the batch and time directions. Following that, the proposed approach’s unbiasedness and convergence are strictly confirmed. Finally, the simulation results for the injection molding process demonstrate that the proposed method is capable of determining the optimal control law as the number of batches is growing increasingly.

Safe and Efficient Take-Off of VTOL UAV Swarms

Currently multicopters are facing a continuous growth in terms of recreational uses, and multiple companies focused on these aircrafts to facilitate certain tasks that were nearly inaccessible to humans, or otherwise involved a great cost. In this context, the drone swarm concept allows us to broaden and incorporate new, more refined applications in which various aircraft coordinate with each other to carry out large-scale tasks. When the number of UAVs involved becomes too high, guaranteeing that the take-off procedure is efficient and yet secure becomes quite complex. Hence, in this paper we propose and validate different algorithms to optimize the take-off time of drones belonging to a swarm, with the objective that there are no collisions between them. In particular, we propose algorithms for both trajectory analysis and batch generation for take-off. Based on a large number of experiments using the ArduSim simulator we prove that the proposed algorithms provide a robust solution within a reasonable time frame when testing with different aerial formations. In addition, we will assess how different UAV position assignment strategies impact our algorithm performance in terms of take-off time and number of batches required.

Exploration of the High-Efficiency Hardware Architecture of SM4-CCM for IoT Applications

The widespread use of the internet of things (IoT) is due to the value of the data collected by IoT devices. These IoT devices generate, process, and exchange large amounts of safety-critical or privacy-sensitive data. Before transmission, the data should be protected against information leakage and data stealing. Deploying authenticated encryption with additional data (AEAD) algorithms on IoT devices ensures data confidentiality and integrity. However, AEAD algorithms are computationally intensive, while IoT devices are resource constrained or even battery powered. Therefore, a low-cost, low-power, and high-efficiency method of implementing an AEAD algorithm into resource-constrained IoT devices is required. The SM4-CCM algorithm, introduced in RFC 8998, is selected as the AEAD algorithm to address this problem. Algorithms similar to SM4-CCM (e.g., SM4 and AES-CCM) provide many architectural design references, but it is challenging to decide which architecture is the most suitable for SM4-CCM. In order to find the most efficient SM4-CCM hardware architecture, a design space exploration method is proposed. Firstly, the SM4-CCM algorithm is divided into five layers, and three candidate architectures are provided for each layer. Secondly, 63 design schemes for SM4-CCM are constructed by combining candidate architectures from each layer. Finally, a batch number of comparisons and analyses of experimental results are used to identify the most efficient one. Under TSMC 90 nm technology, the experimental results of the identified scheme show that the throughput, power consumption, and area achieve 199.99 Mbps, 1.625 mW, and 14.6 K gates, respectively. As a proof of concept, implementing this scheme on an FPGA board is also presented.

Approach of resource expenditure estimation toward mechanization in the manufacturing of cell-based products.

Recent developments for the manufacturing of cell-based products have focused on the advancement of products to clinical trials or commercialization, with awareness of the importance of cost-based effectiveness in cell manufacturing. The mechanization of cell-processing operations is advantageous for the reproducibility and stability of product quality and is thought to reduce the cost-of-goods through the life cycle of the product in a scale-up system; however, few cases of the implementation exist. This study developed an estimation method for the resource expenditure of cell-processing operations in the manufacturing of cell-based products. To estimate resource expenditures, we evaluated the manufacturing processes by operations involving entering into the surrounding area of cell processing zone, materials loading, cell-processing operation, cleaning, and leaving from the surrounding area. The cell-processing operation is applicable to manual or robotic cell manufacturing system in a biosafety cabinet or an isolator system. In cases of low annual batch numbers of manufacturing (batch number <33), the resource expenditure of cell-processing operations in a robotic operation system installed in the isolator system is estimated to be higher compared with a manual operation system in the isolator system due to additional initial costs for design and fabrication of the robotic operation system containing robot arms. With increasing numbers of annual batches, the resource expenditure decreases for robotic operating system, leading to an advantageous juncture where the resource expenditure of a robotic operation system is equivalent to that of a manually operated system, whereby the labor cost for cell-processing operations rises. In addition, the expertise of operations required for cell manufacturing is suggested to foster potential risks associated with the operation skills or turnover of operators, and the cost of education and training increases due to the necessity of persistent human resource development. Collectively, revealing the approach for installation of robotic operation system in cell manufacturing.

Recurrent ligneous conjunctivitis after cataract surgery in a 67-year-old male patient: a case report

BackgroundLigneous conjunctivitis is a rare form of chronic pseudomembranous conjunctivitis which usually starts during infancy. We report on an unsual case of recurrent ligneous conjunctivitis after cataract surgery in a 67-year-old male patient.MethodsThe equipment used for the slit-lamp images was a Haag Streit slit lamp BX900 Sn 00,406 with 16 × magnifications. The used batch number of the camera was sn00406 and the software was from the company CCS Pawlowski Merge Eye. There were no filters used. The images were saved with a resolution of 300 DPI. Neither downstream nor averaging was used to enhance the resolution of the image in the case presentation section or the figure legend.The equipment used for the cross-sectional histologic images was a Zeiss Axioskop 40 microscope with an objective lens Zeiss A-Plan × 20/0.45 (Zoom 6.3 × TV 2/3″″C). The used camera was AxioCam MRc5 and the software was ZEN 3.2. The cross-sectional histologic images were saved with a resolution of 2584 × 1936 Pixels. Neither downstream nor averaging was used to enhance the resolution of the image in the case presentation section or the figure legend.Case presentationThis is a rare case report of ligneous conjunctivitis in a 67-year-old male patient who presented a recurrent conjunctival granuloma after five excisions following cataract surgery in his left eye. We performed a tumor excision with free conjunctival autograft. The histology showed a fibrin crust including macrophages, granulocytes, lymphocytes, and reactively altered squamous cell nests. These findings were consistent with a ″pseudomembrane in conjunctivitis lignosa″. We administered a topical combination of plasminactivator, heparin, cortisone and cyclosporine.ConclusionThis treatment with the combination of plasminactivator, heparin, cortisone and cyclosporine has proven to be effective in preventing the recurrence of ligneous conjunctivitis.

Using a Model-based Material Sparing Approach for Formulation and Process Development of a Roller Compacted Drug Product.

The present work details a material sparing approach that combines material profiling with Instron uniaxial die-punch tester and use of a roller compaction mathematical model to guide both formulation and process development of a roller-compacted drug product. True density, compression profiling, and frictional properties of the pre-blend powders are used as inputs for the predictive roller compaction model, while flow properties, particle size distribution, and assay uniformity of roller compaction granules are used to select formulation composition and ribbon solid fraction. Using less than 10g of a model drug compound for material profiling, roller compacted blend in capsule formulations with appropriate excipient ratios were developed at both 1.4% and 14.4% drug loadings. Subsequently, scale-up batches were successfullymanufactured based on the roller compaction process parameters obtained from predictive modeling. The measured solid fractions of roller compaction ribbon samples from the scale-up batches were in good agreement with the target solid fraction of the modeling. This approach demonstrated considerable advantages through savings in both materials and number of batches in the development of a roller-compacted drug product, which is of particular value at early development stages when drug substance is often limited and timelines are aggressive.

ReComBat: batch-effect removal in large-scale multi-source gene-expression data integration.

With the steadily increasing abundance of omics data produced all over the world under vastly different experimental conditions residing in public databases, a crucial step in many data-driven bioinformatics applications is that of data integration. The challenge of batch-effect removal for entire databases lies in the large number of batches and biological variation, which can result in design matrix singularity. This problem can currently not be solved satisfactorily by any common batch-correction algorithm. We present reComBat, a regularized version of the empirical Bayes method to overcome this limitation and benchmark it against popular approaches for the harmonization of public gene-expression data (both microarray and bulkRNAsq) of the human opportunistic pathogen Pseudomonas aeruginosa. Batch-effects are successfully mitigated while biologically meaningful gene-expression variation is retained. reComBat fills the gap in batch-correction approaches applicable to large-scale, public omics databases and opens up new avenues for data-driven analysis of complex biological processes beyond the scope of a single study. The code is available at https://github.com/BorgwardtLab/reComBat, all data and evaluation code can be found at https://github.com/BorgwardtLab/batchCorrectionPublicData. Supplementary data are available at Bioinformatics Advances online.

A batch scheduling and operator assignment model with time-changing effects to minimise total actual flow time

This paper investigates integrated problems of batch scheduling and operator assignment with time-changing effects due to the operators experiencing different degrees of learning and forgetting, and accumulating different degrees of fatigue. Each batch will be processed through a number of operations in a flow shop, where each operation can be performed by one of the alternative operators. Each operator will be assigned to a maximum of one machine, and a minimum of one operator will be assigned to each machine. The production process could produce a number of defective products that can be reworked. The problems are formulated in a model, and the decision variables are operators' assignment to machines, the number of batches, batch sizes, items allocation to batches, and the schedule of the resulting batches. A proposed algorithm is developed to solve the model, and numerical examples show that the model and the algorithm work effectively for the investigated problem. [Submitted 30 January 2020; Accepted 15 January 2021]

Comparative evaluation of net setting time and radiopacity in Fuji II (GC-Japan) restorative glass ionomer and Iranian glass ionomer

Comparing the net setting time and radiopacity of an Iranian glass ionomer cement (GIC) and Fuji II (GC, Japan) according to ISO 9917-1:2007 standard.In this experimental/in vitro study, for both tests, we prepared 20 samples of Fuji II glass ionomer (self-cure restorative glass ionomer, batch number: 1608031, GC Corporation, Tokyo, Japan) and Iranian glass ionomer (Ava Tajhiz Dandan-Iran) at P/L of 2/7:1. Then, to determine the net setting time, we prepared a metal mold with dimensions of 10 mm in length, 8 mm in width, and 5 mm in height. Ninety seconds after mixing, the surface of the sample was subjected to the indenter, and the net setting time was recorded as the time elapsed between the end of the mixing and the time needle stopped making a complete circular indentation. To determine radiopacity, the specimens were poured into a mold with a diameter of 15 mm and thickness of 1 mm. Samples and a step wedge were irradiated with X-rays. Particle size analysis and Energy-dispersive X-ray spectroscopy (EDS) analysis were also done for both cements. Test results were investigated with SPSS and through independent t-test (P < 0.05).The mean value of net setting time for Fuji II was 4.83 min and for the Iranian Glass ionomer was 3.83 min (P < 0.05). The mean value of radiopacity for Fuji II was 2.3 mmAL and for Iranian Glass ionomer was 1.9 mmAl (P < 0.05).Net setting time and radiopacity of the glass ionomers were within the range of ISO 9917-1:2007. If all properties of the Iranian cement are set appropriately in future investigations, we propose to use it instead of Fuji II GIC. This has the additional benefit of being cost-efficient as Iranian cement costs less than Fuji II cement.

Prediction of per-batch yield rates in production based on maximum likelihood estimation of per-machine yield rates

The demand for high-quality customized products compels manufacturers to adopt batch production. With the ability to accurately estimate batch production yield rates in advance, manufacturers can effectively plan the batch production process and control the production risk based on the estimated values. The per-batch production yield rates can be directly predicted by multiplying the accurately estimated per-machine yield rates corresponding to a batch. Unfortunately, for most manufacturers, the actual per-machine yield rates are difficult to estimate owing to a variety factors. Moreover, per-batch yield-rate prediction has received little attention because recent studies only focused on yield-rate prediction methods for single/continuous production systems. To address this, we propose an expectation-maximization-based approach to predict per-batch yield rates by estimating the per-machine yield rates. Based on the data from T-company, the proposed method could predict the per-batch yield rates for the subsequent week with an average accuracy of 91.86 %, and for five consecutive weeks with an average accuracy of more than 90 %. To further evaluate the performance of the proposed method with different batch production patterns, we conducted simulations to obtain the average accuracy of the estimated per-machine yield rates. In the simulations, the average prediction accuracy of the per-batch yield rates was 91.29 % in the batch production pattern, as in the case of T-company (∼250 machines and ∼1000 batches per week), and it increased as the number of batches increased.

Domain Adaptation and Federated Learning for Ultrasonic Monitoring of Beer Fermentation

Beer fermentation processes are traditionally monitored through sampling and off-line wort density measurements. In-line and on-line sensors would provide real-time data on the fermentation progress whilst minimising human involvement, enabling identification of lagging fermentations or prediction of ethanol production end points. Ultrasonic sensors have previously been used for in-line and on-line fermentation monitoring and are increasingly being combined with machine learning models to interpret the sensor measurements. However, fermentation processes typically last many days and so impose a significant time investment to collect data from a sufficient number of batches for machine learning model training. This expenditure of effort must be multiplied if different fermentation processes must be monitored, such as varying formulations in craft breweries. In this work, three methodologies are evaluated to use previously collected ultrasonic sensor data from laboratory scale fermentations to improve machine learning model accuracy on an industrial scale fermentation process. These methodologies include training models on both domains simultaneously, training models in a federated learning strategy to preserve data privacy, and fine-tuning the best performing models on the industrial scale data. All methodologies provided increased prediction accuracy compared with training based solely on the industrial fermentation data. The federated learning methodology performed best, achieving higher accuracy for 14 out of 16 machine learning tasks compared with the base case model.

A Non-Dominated Genetic Algorithm Based on Decoding Rule of Heat Treatment Equipment Volume and Job Delivery Date

This paper investigated the flexible job-shop scheduling problem with the heat treatment process. To solve this problem, we built an unified mathematical model of the heat treatment process and machining process. Up to now, this problem has not been investigated much. Based on the features of this problem, we are intended to minimize Cmax, maximize the space utilization rate of heat treatment equipment, and minimize the total delay penalty to optimize the scheduling. By taking the dynamic process arrival under consideration, this paper proposed a set of decoding rules based on the heat treatment equipment volume and job delivery date to achieve a hybrid dynamic scheduling solution during one scheduling procedure. When the utilization rate of heat treatment equipment volume is maximized, and the job delivery date is taken under consideration, it is preferred to minimize the number of workpiece batches in the same job, and reduce the waiting time of the pending job. In combination with the improved adaptive non-dominated genetic algorithm, we worked out the solution. Furthermore, we verified the effectiveness of the proposed decoding rules and improved algorithm through algorithm comparison and calculation results. Finally, a software system for algorithm verification and algorithm comparison was developed to verify the validity of our proposed algorithm.