Afet sonrası sahra hastanelerinin yerleşimi için genetik algoritma uygulaması: İstanbul vakası
Afet sonrası sahra hastanelerinin yerleşimi için genetik algoritma uygulaması: İstanbul vakası
Dosyalar
Tarih
2020
Yazarlar
Kömürcü, Yeşim
Süreli Yayın başlığı
Süreli Yayın ISSN
Cilt Başlığı
Yayınevi
Fen Bilimleri Enstitüsü
Özet
Afet sonrası kayıpların çoğu, insani yardım planlamasının olmaması ya da yetersiz uygulamalardan kaynaklanmaktadır. Geçici sahra hastanelerinin yerleşimi ve yaralıların hastanelere atanması doğal afet yönetiminde anahtar konulardır. Mevcut hastanelerin acil servis birimleri bulunmasına rağmen, İstanbul'da ciddi bir deprem olması durumunda bu kapasitelerin yaralılar için yeterli olmayacağı düşünülmektedir. Bu nedenle, felaketin ardından hızla inşa edilecek ve ek kapasite görevi görecek sahra hastanelerine ihtiyaç duyulmaktadır. Bu sahra hastanelerinin en uygun yerlerinin belirlenmesi, yaralıya yanıt verme süresini azaltmak için önemlidir. Ayrıca, yaralıların mevcut hastane ve sahra hastanelerine en uygun şekilde atanması da yanıt süresinin azaltılmasına ve kapasitenin verimli kullanılmasına yardımcı olacaktır. Bu çalışmanın amacı, yaralıların tümüne olabildiğince çabuk yanıt vermek için toplam seyahat maliyetini ve sahra hastanesi kurulum maliyetini en aza indirmektir. Problem NP-Zor türünde olduğundan ve matematiksel modellerin çözümünün çok uzun sürelerde sonuçlanacağı ya da sonuç bulmada yetersiz kalacağından dolayı meta sezgisel yöntemlere başvurulmuştur. Bu amaçla MATLAB'da sezgisel çözüm yöntemi olan genetik algoritma (GA) geliştirilmiştir. Algoritmanın performansını artırmak için farklı çaprazlama ve yer değiştirme stratejileri test edilmiştir. Deneysel çalışmada, 4 farklı GA stratejilerinin performansları, optimal sonucunun bilindiği deneysel veri kümeleri kullanılarak karşılaştırılmıştır. Çaprazlama tiplerinden birleşim çaprazlama, iki noktalı çaprazlamaya göre daha iyi performans göstermiş ve optimal sonucu bulmuştur. Yer değiştirme tiplerinde ise en kötü bireyi eleyerek yeni jenerasyon oluşturan algoritma popülasyonun %50'sini eleyerek yeni jenerasyon oluşturan algoritmadan daha kısa sürede optimal sonuca ulaşmıştır. Karşılaştırma sonucuna göre, gerçek İstanbul veri seti için en iyi GA seçilmiş ve gerçek İstanbul verisinde uygulanmıştır. Japonya İş Birliği Uluslararası Ajansı'na göre olası yıkıcı bir İstanbul depreminde en fazla ölü ve ağır yaralı olacak ilçelerden Bahçelievler ve Küçükçekmece uygulama için seçilmiştir. Uygulamada Bahçelievler ve Küçükçekmece ilçelerinde mesafe kısıtı olmaksızın ve mesafe kısıtı eklenerek 2 model çözülmektedir. Duyarlılık analizi kapsamında farklı yaralı sayıları, mesafe kısıtı ve sahra hastanesi kapasitesi ile oluşturulan 8 model çoklu koşumlar sonucunda değerlendirilmiştir. Yaralı sayısı ve sahra hastanesi kurulum maliyeti sabit olan mesafe kısıtının değiştiği modeller incelendiğinde mesafe kısıtı yarıçapı azaldıkça açılan sahra hastanesi sayısı arttığından seyahat maliyeti azalsa da toplam maliyet artmaktadır. Yaralı sayısı ve mesafe kısıtı aynı olup sahra hastanesi kapasitesi ve dolayısıyla kurulum maliyeti arttığında ise açılan sahra hastanesi sayısı oldukça azalmaktadır. Sahra hastanesi kapasitesi ve mesafe kısıtı aynı olup yaralı sayısı değiştiğinde yaralı sayısının artışına göre seyahat maliyeti ve kurulum maliyeti doğru orantılı olarak artmaktadır. Model 4 en az maliyetli model olup Model 4'ün özelliklerine bakıldığında; yaralılar en fazla 5 km yarıçapında bulunan hastanelere atanmaktadır, hem maliyet daha az hem de yaralıların seyahat süresi oldukça kısalmış olmaktadır.
Natural disasters affect the lives of human beings all around the world. Moreover, they cause big losses. Most of the post-disaster losses occur due to the lack of humanitarian relief planning or inadequate implementations. Timely and effective response to disaster required the interaction and coordination of many parties. After the natural disaster such as earthquake, fire, etc. the medical intervention for the injured people should be given as quick as possible. The capacities of existing emergency units of the existing hospitals are expected to be not enough for the injured people in case of a serious earthquake in Istanbul. This study deals with the post-disaster relief planning for the expected earthquake in Istanbul which is located on a major fault line and hosts more than 17 million populations. The field hospitals which will be quickly constructed after the disaster aim to serve as additional capacity needed. According to the World Health Organization, field hospital: A mobile, self-contained, self-sufficient health care facility capable of rapid deployment and expansion or contraction to meet immediate emergency requirements for a specified period of time. Field hospital, its services include an operating theater, intensive observation, anaesthesia, x-ray, laboratory, maternal-child health, pharmacy, sterilization, outpatient clinics. Provides safe medical and surgical interventions, while offering limited medical/surgical care. The problem we study is finding the optimum locations of the field hospitals considering the existing hospitals and the assignment of the injured people to the hospitals. A mixed integer model is formulated for the problem. The objectives of the model are to minimize the total travel cost and the total cost of establishing field hospitals in order to respond to all of the injured as quickly as possible. Since, finding the optimal result for the real-sized problems is not possible with the mixed integer programming, a heuristic solution methodology, GA is developed in MATLAB. Different crossover and replacement strategies have been tested in order to improve the performance of the algorithm. In the experimental study, the performances of different GAs strategies are compared using experimental datasets where optimal results are known. The four different GAs are run for several times and the averages of the obtained results show that two point crossover do not give the optimal result. Thus, we eliminate the use of two point crossover and continue to use the union crossover by which the optimal solution is obtained. Using the union crossover, the replacement strategies are also differentiated as worst 50% replacement and worst 1 replacement. When these two replacement strategies are compared both give the optimal result but the replacement of 1 population finds faster the optimal solution. According to earthquake experts a strong earthquake is expected in the near future. There exist pre-disaster preparations for the incident. Also anticipations related to post-disaster situation are also generated and post-disaster planning is made. In our case study, we aim to find the optimal field hospital locations and patient locations to them in two districts of Istanbul, namely Bahcelievler and Kucukcekmece which are two large districts is expected to be severely affected by an earthquake. Many regulations have been taken into account when creating the data. During the disaster, the multi-storey structure of the existing hospitals cannot be used for security reasons. Only emergency departments of hospitals can be used. The capacity of the emergency units of the existing hospitals is related to the number of beds in the hospital. Field hospitals can be established in open areas such as school gardens, mosque courtyards, playgrounds, football field carpet in a region. In the study, school gardens, parks and football field carpet areas in districts will be used for potential field hospitals. The size of a field hospital was determined as 520 m2. The capacity of a field hospital was determined to be 208 people. In the computational study, we conduct two sets of runs as follows: (i) without a travel distance limit for the assignment of patients to the hospitals, (ii) with a travel distance limit for the assignment of patients to the hospitals. Then we compare these results in terms of travel costs arising from the patient-hospital assignment and the setup costs of the field hospitals. In the Model 1 (without a travel distance limit for the assignment of patients to the hospitals), the best chromosome (due to minimum total cost) is shown 38 of patient location is assigned all of existing hospitals. 106 field hospitals were opened for others. The fitness function value consists of 94.95% setup cost and 5.05% travel cost. 44% of field hospitals is opened. Average unutilized capacity is 55 person/hospital. Unutilized capacity of hospitals is safety margin for unforeseen situation. In the Model 2 (with a travel distance limit for the assignment of patients to the hospitals), the best chromosome (due to minimum total cost) is shown 36 of patient locations is assigned all of existing hospitals. 107 field hospitals were opened for others. The fitness function value consists of 95.4% setup cost and 4.6% travel cost. 44% of field hospitals is opened. Average unutilized capacity is 57 person/hospital. Unutilized capacity of hospitals is safety margin for unforeseen situation. When distance limit is added, it is observed that setup cost increased and travel cost decreased. Within the scope of sensitivity analysis, 8 models occured with different injured numbers, distance constraints and field hospital capacity. The results obtained are as follows; The total cost increases as the distance constraint radius decreases. When the field hospital capacity and therefore the cost of installation increases, the number of field hospitals opened is considerably reduced. When the number of injured changes, travel cost and installation cost increase proportionally according to the increase in the number of injured. Between 10 models, Model 4 is the least costly model. Looking at the features of Model 4; the injured are assigned to hospitals with a maximum radius of 5 km, both the cost is less and the travel time of the injured is considerably shortened. In the Model 4, the best chromosome is shown 35 of patient location is assigned all of existing hospitals. 53 field hospitals were opened for others. The fitness function value consists of 95.2% setup cost and 4.8% travel cost. 22% of field hospitals is opened. In the future research, factors affecting the travel (such as damage to the road, demolition of the building to the road) can be taken into account. Moreover, the supply transportation to the hospitals from specified warehouses can also be included in the location model. The performance of the genetic algorithm can be compared with different meta heuristic methods. Unused capacity can be reduced by assigning injured locations to more than one hospital rather than a hospital.
Natural disasters affect the lives of human beings all around the world. Moreover, they cause big losses. Most of the post-disaster losses occur due to the lack of humanitarian relief planning or inadequate implementations. Timely and effective response to disaster required the interaction and coordination of many parties. After the natural disaster such as earthquake, fire, etc. the medical intervention for the injured people should be given as quick as possible. The capacities of existing emergency units of the existing hospitals are expected to be not enough for the injured people in case of a serious earthquake in Istanbul. This study deals with the post-disaster relief planning for the expected earthquake in Istanbul which is located on a major fault line and hosts more than 17 million populations. The field hospitals which will be quickly constructed after the disaster aim to serve as additional capacity needed. According to the World Health Organization, field hospital: A mobile, self-contained, self-sufficient health care facility capable of rapid deployment and expansion or contraction to meet immediate emergency requirements for a specified period of time. Field hospital, its services include an operating theater, intensive observation, anaesthesia, x-ray, laboratory, maternal-child health, pharmacy, sterilization, outpatient clinics. Provides safe medical and surgical interventions, while offering limited medical/surgical care. The problem we study is finding the optimum locations of the field hospitals considering the existing hospitals and the assignment of the injured people to the hospitals. A mixed integer model is formulated for the problem. The objectives of the model are to minimize the total travel cost and the total cost of establishing field hospitals in order to respond to all of the injured as quickly as possible. Since, finding the optimal result for the real-sized problems is not possible with the mixed integer programming, a heuristic solution methodology, GA is developed in MATLAB. Different crossover and replacement strategies have been tested in order to improve the performance of the algorithm. In the experimental study, the performances of different GAs strategies are compared using experimental datasets where optimal results are known. The four different GAs are run for several times and the averages of the obtained results show that two point crossover do not give the optimal result. Thus, we eliminate the use of two point crossover and continue to use the union crossover by which the optimal solution is obtained. Using the union crossover, the replacement strategies are also differentiated as worst 50% replacement and worst 1 replacement. When these two replacement strategies are compared both give the optimal result but the replacement of 1 population finds faster the optimal solution. According to earthquake experts a strong earthquake is expected in the near future. There exist pre-disaster preparations for the incident. Also anticipations related to post-disaster situation are also generated and post-disaster planning is made. In our case study, we aim to find the optimal field hospital locations and patient locations to them in two districts of Istanbul, namely Bahcelievler and Kucukcekmece which are two large districts is expected to be severely affected by an earthquake. Many regulations have been taken into account when creating the data. During the disaster, the multi-storey structure of the existing hospitals cannot be used for security reasons. Only emergency departments of hospitals can be used. The capacity of the emergency units of the existing hospitals is related to the number of beds in the hospital. Field hospitals can be established in open areas such as school gardens, mosque courtyards, playgrounds, football field carpet in a region. In the study, school gardens, parks and football field carpet areas in districts will be used for potential field hospitals. The size of a field hospital was determined as 520 m2. The capacity of a field hospital was determined to be 208 people. In the computational study, we conduct two sets of runs as follows: (i) without a travel distance limit for the assignment of patients to the hospitals, (ii) with a travel distance limit for the assignment of patients to the hospitals. Then we compare these results in terms of travel costs arising from the patient-hospital assignment and the setup costs of the field hospitals. In the Model 1 (without a travel distance limit for the assignment of patients to the hospitals), the best chromosome (due to minimum total cost) is shown 38 of patient location is assigned all of existing hospitals. 106 field hospitals were opened for others. The fitness function value consists of 94.95% setup cost and 5.05% travel cost. 44% of field hospitals is opened. Average unutilized capacity is 55 person/hospital. Unutilized capacity of hospitals is safety margin for unforeseen situation. In the Model 2 (with a travel distance limit for the assignment of patients to the hospitals), the best chromosome (due to minimum total cost) is shown 36 of patient locations is assigned all of existing hospitals. 107 field hospitals were opened for others. The fitness function value consists of 95.4% setup cost and 4.6% travel cost. 44% of field hospitals is opened. Average unutilized capacity is 57 person/hospital. Unutilized capacity of hospitals is safety margin for unforeseen situation. When distance limit is added, it is observed that setup cost increased and travel cost decreased. Within the scope of sensitivity analysis, 8 models occured with different injured numbers, distance constraints and field hospital capacity. The results obtained are as follows; The total cost increases as the distance constraint radius decreases. When the field hospital capacity and therefore the cost of installation increases, the number of field hospitals opened is considerably reduced. When the number of injured changes, travel cost and installation cost increase proportionally according to the increase in the number of injured. Between 10 models, Model 4 is the least costly model. Looking at the features of Model 4; the injured are assigned to hospitals with a maximum radius of 5 km, both the cost is less and the travel time of the injured is considerably shortened. In the Model 4, the best chromosome is shown 35 of patient location is assigned all of existing hospitals. 53 field hospitals were opened for others. The fitness function value consists of 95.2% setup cost and 4.8% travel cost. 22% of field hospitals is opened. In the future research, factors affecting the travel (such as damage to the road, demolition of the building to the road) can be taken into account. Moreover, the supply transportation to the hospitals from specified warehouses can also be included in the location model. The performance of the genetic algorithm can be compared with different meta heuristic methods. Unused capacity can be reduced by assigning injured locations to more than one hospital rather than a hospital.
Açıklama
Tez (Yüksek Lisans) -- İstanbul Teknik Üniversitesi, Fen Bilimleri Enstitüsü, 2020
Anahtar kelimeler
metasezgiseller,
metaheuristics,
sahra-acil durum hastaneleri,
emergency-field hospitals