State estimation for anaerobic digesters using the ADM1.

The optimization of full-scale biogas plant operation is of great importance to make biomass a competitive source of renewable energy. The implementation of innovative control and optimization algorithms, such as Nonlinear Model Predictive Control, requires an online estimation of operating states of biogas plants. This state estimation allows for optimal control and operating decisions according to the actual state of a plant. In this paper such a state estimator is developed using a calibrated simulation model of a full-scale biogas plant, which is based on the Anaerobic Digestion Model No.1. The use of advanced pattern recognition methods shows that model states can be predicted from basic online measurements such as biogas production, CH4 and CO2 content in the biogas, pH value and substrate feed volume of known substrates. The machine learning methods used are trained and evaluated using synthetic data created with the biogas plant model simulating over a wide range of possible plant operating regions. Results show that the operating state vector of the modelled anaerobic digestion process can be predicted with an overall accuracy of about 90%. This facilitates the application of state-based optimization and control algorithms on full-scale biogas plants and therefore fosters the production of eco-friendly energy from biomass.

Impact of healthcare strategies on patterns of paediatric sight impairment in a developed population: 1984-2011.

PurposeThe aim of our study was to analyse paediatric sight-impairment trends in Northern Ireland (NI) over a 28-year period to better understand the impact which changes in health-care provision may be having on childhood blindness and to enable us to assess our progress towards achieving the World Health Organisation (WHO) aims.MethodsA database of Certificates of Visual Impairment completed for NI children aged <16 years was used to determine the cause of sight impairment from 1984 to 2011. Causes were classified into preventable or treatable conditions and analysed for trends.ResultsFive hundred and ninety-eight children were registered as having impaired vision over the 28-year period. A total of 22% had preventable or treatable conditions. Optic atrophy was the most common cause responsible for 16% of registrations followed by albinism (12%), cerebral visual impairment (11%), congenital cataract (8%), retinopathy of prematurity (ROP) (8%), and congenital motor nystagmus (7%). The incident rate (per million population aged <16 years) for registerable vision loss due to congenital cataract decreased from 5.89 (CI 2.82-10.83) in 1984-1987 to 2.63 (CI 0.72-6.74) in 2008-2011. For ROP, the incident rate peaked during 2000-2003 at 8.87 (CI 4.85-14.88). Thereafter, there was a statistically significant reduction in incident rate to 1.98 (CI 0.41-5.77) in 2008-2011 (P=0.008).ConclusionsSight-impairment registrations due to preventable or treatable causes have decreased over the past 28 years. This is likely due to better surgical techniques and improved refractive care for conditions such as congenital cataract and ongoing research and treatment protocols for conditions such as ROP. Future advances in this area may help to further reduce the burden of childhood sight impairment and improve quality of life for these patients.

Long term functional and structural outcomes of laser therapy for retinopathy of prematurity.

AIMS: To assess the long term functional and structural outcomes of premature babies who received diode laser photocoagulation for threshold retinopathy of prematurity (ROP). METHODS: 25 patients (43 eyes) treated with laser were recalled for assessment at a mean follow up of 11 years. A further seven patients (14 eyes) with subthreshold ROP, which had regressed spontaneously without laser treatment, were also examined. All children underwent distance acuity, near acuity, contrast sensitivity (CS), and colour vision assessments followed by a dilated fundal examination and cycloplegic autorefraction. RESULTS: The laser treated eyes had a mean distance visual acuity of 0.37 logMAR, a mean near visual acuity of 0.39 logMAR, a mean contrast sensitivity of 1.49 log CS units, and a mean spherical equivalent of -2.10D. An unfavourable distance visual acuity outcome occurred in five eyes (13.5%). An unfavourable near visual acuity outcome was also noted in the five eyes (13.5%) with poor distance visual outcome. 7% had an unfavourable structural outcome. On comparison with the control group, there was no significant difference in near acuity, CS, refraction, or colour vision between the two groups. However, there was a statistically significant difference in terms of distance visual acuity (p = 0.03). CONCLUSIONS: Laser treated eyes with favourable structural outcome have a good visual outcome. The results show a long term benefit from diode laser photocoagulation in preserving distance and near vision in eyes with threshold ROP.

RetCam image analysis of optic disc morphology in premature infants and its relation to ischaemic brain injury.

AIMS: To assess optic disc characteristics in premature infants with and without ischaemic brain injury and to evaluate the role of optic disc morphology in dating the injury. METHODS: RetCam fundal images, cranial ultrasounds and magnetic resonance imaging (MRI) of 109 premature infants were analysed. The study cohort was divided into subgroups depending on the presence or absence of periventricular leucomalacia (PVL) and intraventricular haemorrhage (IVH). The control group consisted of infants with normal neuroimaging at term and 2 years of age. Using the image analysis software of the RetCam, optic disc diameter (ODD), optic disc area (ODA), and optic cup area (OCA) were measured at 33-34 weeks gestational age. As serial cranial ultrasonography had been performed, it was possible to date the brain injury in those infants with periventricular white matter (PVWM) damage. RESULTS: Although there was a trend towards reducing ODD, ODA, and OCA with increasing severity of IVH, only the IVH 4 group differed significantly from the controls for these parameters (p = 0.002, p = 0.02, and p = 0.04, respectively). 44.4% of infants with grade 4 IVH had small discs. Only one patient had a large cup in a normal sized disc; this patient had IVH 4. In patients with PVWM damage, the median time of insult was 27 weeks in those with small discs and 28 weeks in those with normal discs. This difference was not significant (p = 0.23). CONCLUSIONS: Premature infants with IVH 4 have an increased incidence of optic nerve hypoplasia. We found no association between disc morphology and timing of brain injury.

Fast parallel off-line training of multilayer perceptrons.

Various approaches to the parallel implementation of second-order gradient-based multilayer perceptron training algorithms are described. Two main classes of algorithm are defined involving Hessian and conjugate gradient-based methods. The limited- and full-memory Broyden-Fletcher-Goldfarb-Shanno (BFGS) algorithms are selected as representative examples and used to show that the step size and gradient calculations are critical components. For larger problems the matrix calculations in the full-memory algorithm are also significant. Various strategies are considered for parallelization, the best of which is implemented on parallel virtual machine (PVM) and transputer-based architectures. Results from a range of problems are used to demonstrate the performance achievable with each architecture. The transputer implementation is found to give excellent speed-ups but the problem size is limited by memory constraints. The speed-ups achievable with the PVM implementation are much poorer because of inefficient communication, but memory is not a difficulty.

A hybrid linear/nonlinear training algorithm for feedforward neural networks.

This paper presents a new hybrid optimization strategy for training feedforward neural networks. The algorithm combines gradient-based optimization of nonlinear weights with singular value decomposition (SVD) computation of linear weights in one integrated routine. It is described for the multilayer perceptron (MLP) and radial basis function (RBF) networks and then extended to the local model network (LMN), a new feedforward structure in which a global nonlinear model is constructed from a set of locally valid submodels. Simulation results are presented demonstrating the superiority of the new hybrid training scheme compared to second-order gradient methods. It is particularly effective for the LMN architecture where the linear to nonlinear parameter ratio is large.